allMathematicsPhysicsEngineeringBioinformaticsMedical SciencesSocial SciencesEarth SciencesComputer ScienceBiologyFinanceEconomicsChemistry

All

Associated with article

[Computer Science\ \ 11 | Jun | 2026\ \ QRTlib: A Library for Fast Quantum Real Transforms\ \ QRTlib \ \ This library implements Quantum Real Transforms (QRTs), including:\ \

• Quantum Hartley Transform (QHT)\ \
  • Approach 1: Recursive QHT proposed in [1].\
  • Approach 2: Using Linear Combination of Unitaries (LCU) proposed in [2].\
• Quantum Sine and Cosine Transforms\ \
• Approach: Optimized implementation of [3].\ \
• Types:\ \
  • Quantum Sine and Cosine Transform type I\
  • Quantum Sine and Cosine Transform type II\
  • Quantum Sine and Cosine Transform type III\
  • Quantum Sine and Cosine Transform type IV\ \

• • *\ \ Installation \ \ To install dependencies:\ \ bash\ \ Copy\ \

pip install -r requirements.txt\\
```\\
\\
**Usage example** \\
\\
The following snippet adds Quantum Hartley Transform (QHT) to your circuit using LCU method:\\
\\
- _Notes:_ For recursive method one can use (`type="REC"`).\\
\\
python\\
\\
Copy\\
\\
```python\\
from qiskit import QuantumCircuit\\
from QHTGate import QHTGate\\
\\
# Example for quantum hartley transform of using LCU method for 4 qubits\\
gate = QHTGate(4, type="LCU")\\
qc = QuantumCircuit(8)\\
qc.append(gate, list(range(8)))\\
```\\
\\
**References**\\
\[1\] Doliskani, Jake and Mirzaei, Morteza and Mousavi, Ali. "Public-key quantum money and fast real transforms". (2025)\\
\[2\] Ahmadkhaniha, Armin and Doliskani, Jake and Chen, Lu and Sun, Zhifu. "QRTlib: A Library for Fast Quantum Real Transforms". (2025)\\
\[3\] Klappenecker, Andreas and Rotteler, Martin. "Discrete cosine transforms on quantum computers". (2001)\\
\\
Armin Ahmadkhaniha, Lu Chen, Jake Doliskani & Zhifu Sun](/content/explore/4bd142b4-5f0a-4081-8df1-52b32ad3b58a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3553607/tree/v1/index.html)

[Computer Science\\
\\
10 \| Jun \| 2026\\
\\
From Speech to Cognitive Map: An LLM-Based Framework for SODA-Driven Stakeholder Needs Analysis in Systems-of-Systems\\
\\
This artifact supports the reproducibility of a research pipeline that applies a Large Language Model (LLM) to automate the generation of SODA (Strategic Options Development and Analysis) cognitive maps from semi-structured interview transcripts. The study is grounded in Design Science Research (DSR) and applied to the Brazilian Unmanned Traffic Management system (BR-UTM) as an empirical case of System-of-Systems (SoS) integration.\\
\\
Francisco Henrique Figueiredo Araujo, Christopher Shneider Cerqueira & Guilherme Micheli Bedini Moreira](/content/explore/a315d134-8e01-41ea-9220-b0b0a368cb05?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9910105/tree/v1/index.html)

[Computer Science\\
\\
10 \| Jun \| 2026\\
\\
EEG-learning: an integrate Python framework for experimental studies applying causal connec- tivity inference, preprocessing and spectral-temporal analysis\\
\\
A complete pipeline to preprocess, quantify, and visualize EEG power in studies with customizable sessions and experimental conditions. Inputs follow BIDS; outputs are publication-ready CSVs and figures. The included data are provided for demonstration purposes only and should not be interpreted as real experimental results.\\
\\
Paloma Alves et al.](/content/explore/ccefd871-dc47-43f1-944f-72f79d8a3483?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3934088/tree/v1/index.html)

[Economics\\
\\
9 \| Jun \| 2026\\
\\
Forecasting daily visits in Shanghai with Model Combination and Telco Big Data\\
\\
Code to reproduce the forecasts in the paper "Forecasting daily visits in Shanghai with Model Combination and Telco Big Data".\\
\\
Giulia CARALLO, Roberto Casarin, Nicola Camatti & Xiang Feng](/content/explore/d7b48482-cc3b-4471-97c0-083ebf49978c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7490765/tree/v1/index.html)

[Computer Science\\
\\
9 \| Jun \| 2026\\
\\
Probabilistic multi-label classification via a divide-and-conquer and fusion approach\\
\\
Reproduces the CHD-49 target × evaluation table from the paper (PCC + Logistic Regression, 10-fold CV, seeds 1–5). CPU-only, finishes in seconds. Output: result\_CHD\_49\_crosstab.csv, in which every diagonal entry (target = evaluation metric) is the maximum of its column — the paper's central claim.\\
\\
Vu-Linh Nguyen](/content/explore/8ee6157e-7dac-43ff-bbde-8e2e0396b3c0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7245746/tree/v1/index.html)

[Associated article](https://doi.org/10.1016/j.inffus.2026.104517) published in [Information Fusion](/content/explore?query=Information%20Fusion&refine=journal/index.html)

[Engineering\\
\\
9 \| Jun \| 2026\\
\\
Online Learning for Supervisory Switching Control (code)\\
\\
Source code for paper "Online Learning for Supervisory Switching Control"\\
\\
Haoyuan Sun & Ali Jadbabaie](/content/explore/e018709c-0afa-4d93-831f-0da416e066fa?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1890709/tree/v1/index.html)

[Computer Science\\
\\
9 \| Jun \| 2026\\
\\
A Triple-Layer Framework for Programmable Asset-Level Participation\\
\\
Reference Sui Move prototype accompanying the manuscript "A Triple-Layer Framework for Programmable Asset-Level Participation".\\
\\
The artifact illustrates Layer-3 settlement semantics and oracle-prepared monthly distribution logic for Programmable Rights Tokens (PRTs).\\
\\
It is provided as an implementation-alignment artifact supporting the architectural concepts presented in the manuscript.\\
\\
Umut Pekel](/content/explore/9738094b-ca17-448f-be3d-38993f5a6afb?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9502146/tree/v1/index.html)

[Computer Science\\
\\
8 \| Jun \| 2026\\
\\
Version 2 - Fumeus: A family of Python tools for text mining with smoke terms\\
\\
Fumeus is a family of Python tools for performing smoke term analyses. "Smoke terms" were first popularized in safety surveillance as sets of words and phrases particularly indicative of safety hazard-related language. More broadly, these methods can be utilized to develop sets of words and phrases associated with a dependent variable and to rank or score text to surveil for that dependent variable.\\
\\
David M. Goldberg, Richard J. Gruss & Alan S. Abrahams](/content/explore/e0cc86ad-ccf4-4c79-a54c-eef31eacbc45?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5942556/tree/v1/index.html)

[Engineering\\
\\
8 \| Jun \| 2026\\
\\
On the Distribution of a Two-Dimensional Random Walk with Restricted Angles](/content/explore/a272bcbc-7e0a-452a-88d2-00c3c639e9dd?page=2&filter=all/index.html)

[This repository is accompanying the paper "On the Distribution of a Two-Dimensional Random Walk with Restricted Angles" (Karl-Ludwig Besser, IEEE Transactions on Signal Processing, 2026.](/content/explore/a272bcbc-7e0a-452a-88d2-00c3c639e9dd?page=2&filter=all/index.html) [DOI:10.1109/TSP.2026.3694607](https://doi.org/10.1109/TSP.2026.3694607), [arXiv:2507.15475](https://arxiv.org/abs/2507.15475)).

In this paper, we derive the distribution of a two-dimensional (complex) random walk in which the angle of each step is restricted to a subset of the circle. This setting appears in various domains, such as in over-the-air computation in signal processing. In particular, we derive the exact joint and marginal distributions for two steps, numerical solutions for a general number of steps, and approximations for a large number of steps. Furthermore, we provide an exact characterization of the support for an arbitrary number of steps. The results in this work provide a reference for future work involving such problems.

Karl-Ludwig Besser

[Open Capsule](/content/capsule/3825776/tree/v1/index.html)

[Engineering\\
\\
8 \| Jun \| 2026\\
\\
Reproducibility Package for Predictive STAR-RIS/FAS Near-Field ISAC\\
\\
This capsule contains the reproducibility code and data associated with the manuscript “Predictive STAR-RIS and Fluid Antenna Configuration for Near-Field ISAC Under Blockage Uncertainty”. It regenerates the numerical figures and verifies the ablation/outage table values reported in the manuscript.\\
\\
El Miloud Ar-Reyouchi, Saadia Chabel & Kamal Ghoumid](/content/explore/5ae77f64-1af0-4df7-b817-5ba3001fa37f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3457237/tree/v1/index.html)

[Physics\\
\\
8 \| Jun \| 2026\\
\\
New Insights from TRMM Legacy: Stage-Dependent Relationships Between Lightning and Convective Structure in Thunderstorms Over Tropical Africa](/content/explore/12111b7d-5a49-4970-9020-e13307a30ef3?page=2&filter=all/index.html)

[You can get more information in My GitHub and Zenodo](/content/explore/12111b7d-5a49-4970-9020-e13307a30ef3?page=2&filter=all/index.html)

[Github:](/content/explore/12111b7d-5a49-4970-9020-e13307a30ef3?page=2&filter=all/index.html) [https://github.com/oldsixxiaolv/Manuscript\_Lvyh](https://github.com/oldsixxiaolv/Manuscript_Lvyh)
Zenodo: [https://doi.org/10.5281/zenodo.18041880](https://doi.org/10.5281/zenodo.18041880)

Yihang Lyu

[Open Capsule](/content/capsule/5084350/tree/v1/index.html)

[Associated article](https://doi.org/10.1029/2026jd046836) published in [Journal of Geophysical Research: Atmospheres](/content/explore?query=Journal%20of%20Geophysical%20Research%3A%20Atmospheres&refine=journal/index.html)

[Biology\\
\\
8 \| Jun \| 2026\\
\\
pycasa: A Python toolkit for computer-assisted semen analysis](/content/explore/8be8c5d6-4ce6-4125-a7cf-a85bf310eb15?page=2&filter=all/index.html)

[**pycasa** \\
**Version 0.0.1**\\
\\
`pycasa` is a Python toolkit for computer-assisted semen analysis workflows. It supports loading microscopy videos, preprocessing frames, running detection and tracking, computing motility metrics, assessing predictions against groundtruth, and visualizing results.\\
**Why pycasa**\\
\\
- Fluent session API centered on a single `Casa` object.\\
- Modular namespaces for each workflow stage.\\
- Support for both default reference data and custom video pipelines.\\
\\
**Quick Install**\\
\\
bash\\
\\
Copy\\
\\
```bash\\
pip install "git+https://github.com/DFL-KamLab/pycasa.git"\\
```\\
\\
For the full default-data + YOLO example:\\
\\
bash\\
\\
Copy\\
\\
```bash\\
pip install "pycasa[io,yolo] @ git+https://github.com/DFL-KamLab/pycasa.git"\\
```\\
\\
**Starter Example**\\
\\
python\\
\\
Copy\\
\\
```python\\
import pycasa as pc\\
\\
self = pc.io.load_default_data()\\
self.preprocessing.binarization.otsu()\\
self.detection.yolo()\\
self.info()\\
```\\
\\
**Package Structure**\\
\\
text\\
\\
Copy\\
\\
```text\\
pycasa/\\
  casa/            # Fluent wrappers (self.io, self.detection, ...)\\
  io/              # Video/default-data loading implementations\\
  preprocessing/   # Grayscale, normalization, binarization implementations\\
  detection/       # Detection backends (moving-cells, digital washing, Urbano, YOLO v5/v26)\\
  tracking/        # Tracking backends (SORT, DeepSORT, JPDAF)\\
  motility/        # Motility parameter computation\\
  assessment/      # Prediction-vs-groundtruth evaluation\\
  visualization/   # Plotting and interactive analysis\\
  _core/           # Session schema/validation primitives\\
  utils/           # Shared helper utilities\\
```\\
\\
**Documentation Website** \\
For detailed setup, examples, and API references, use the website:](/content/explore/8be8c5d6-4ce6-4125-a7cf-a85bf310eb15?page=2&filter=all/index.html)

- [Home:](/content/explore/8be8c5d6-4ce6-4125-a7cf-a85bf310eb15?page=2&filter=all/index.html) [https://dfl-kamlab.github.io/pycasa/](https://dfl-kamlab.github.io/pycasa/)
- Setup & Requirements: [https://dfl-kamlab.github.io/pycasa/getting-started/setup/](https://dfl-kamlab.github.io/pycasa/getting-started/setup/)
- Examples: [https://dfl-kamlab.github.io/pycasa/examples/default-data-otsu-yolo/](https://dfl-kamlab.github.io/pycasa/examples/default-data-otsu-yolo/)
- API Guide: [https://dfl-kamlab.github.io/pycasa/api/casa-session/](https://dfl-kamlab.github.io/pycasa/api/casa-session/)

Atilla Sivri, Moshe Kam & Ludvik Alkhoury

[Open Capsule](/content/capsule/3069672/tree/v1/index.html)

[Computer Science\\
\\
8 \| Jun \| 2026\\
\\
A DNA-Inspired Cryptographic Framework for Enterprise Network Security Simulation\\
\\
A DNA-Inspired Cryptographic Framework for\\
Enterprise Network Security Simulation\\
\\
Ernest King Arthur](/content/explore/e1003134-3d37-41f1-aa19-c13e159a2c19?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3620383/tree/v1/index.html)

[Computer Science\\
\\
5 \| Jun \| 2026\\
\\
Enhanced Chimera\\
\\
This repository contains a comprehensive, MATLAB-based end-to-end baseband simulator for the Enhanced Chimera protocol. It strictly evaluates the physical-layer robustness and cryptographic integrity of the GPS L1C (CNAV-2) signal utilizing Deterministic Location, Cryptographic Value (DLCV) puncturing.\\
\\
Unlike conventional implementations, this simulator uniquely integrates rigorous cryptographic authentication directly into the physical layer signal processing chain. The simulation demonstrates that a 900-bit cryptographic payload can be seamlessly embedded, transmitted over a noisy channel, naturally tracked by a Phase-Locked Loop (PLL), and authenticated via ECDSA-512 and TESLA hash chains without compromising the backward compatibility of the Standard Chimera protocol.\\
\\
Chang-Seop Park](/content/explore/0508c19d-5f33-4732-ab43-18c3dcdeb660?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4847360/tree/v1/index.html)

[Computer Science\\
\\
8 \| Jun \| 2026\\
\\
Detecting Money Laundering in Bank Deposit Transactions\\
\\
Detecting Money Laundering in Bank Deposit Transactions\\
\\
C. S. Souza](/content/explore/6b2ad3db-03c0-4aa3-a71a-e42d2e886335?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8312751/tree/v1/index.html)

[Computer Science\\
\\
5 \| Jun \| 2026\\
\\
ATTEST: Post-Quantum Verifiable Credential Revocation for IoT Networks\\
\\
ATTEST is a post-quantum revocation framework for IoT verifiable credentials, combining the COMPASS lattice accumulator with a Bloom-filter layer.\\
The accumulator proves credential issuance; the Bloom filter enforces current validity. Constant-size epoch updates (4 KB add, 16 KB revoke) satisfy mass and offline revocation definitions under V-SIS hardness assumptions.\\
\\
Tao-hsiang Chang, Masahiro Mambo & Raylin Tso](/content/explore/9e30d158-62c8-4a48-92ad-ea6dcb5b4b03?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8090555/tree/v1/index.html)

[Computer Science\\
\\
5 \| Jun \| 2026\\
\\
Learning Beyond Empiricism with Introspective Optimization via Statistical Complexity\\
\\
Traditional deep learning relies solely on data to adjust model parameters, consistently with empiricist thinking. However, empiricism does not account for theory-ladenness, which can only be addressed through an internal pattern shift. Although theory-ladenness manifests as overfitting, no method has been proposed to shift patterns and mitigate it. We introduce a introspective learning approach that combines internal statistical complexity with a traditional data-based loss function. The method was evaluated on a decoder-only Transformer trained on the WikiText-2 training split, using its official validation split for monitoring, and tested on both the WikiText-2 test split (in-distribution) and the Tiny Shakespeare dataset (out-of-distribution generalization). Results show a 38.56% drop in final loss when using databases other than the training set, evidencing significantly greater generalization. Introspective learning improves model generalization by optimizing internal statistical complexity, potentially mitigating data over-specialization in language models.\\
\\
Lucas Miranda Mendonça Rezende, Zhao Liang & Luiz Otávio Murta Jr.](/content/explore/a2039e33-3f3b-4c8e-afcc-bc0744ff5910?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0818876/tree/v1/index.html)

[Bioinformatics\\
\\
4 \| Jun \| 2026\\
\\
ILT2 identifies an unexploited pool of intratumoral CD8+ bystander T cells with TCR-independent cytotoxicity in renal cell carcinoma\\
\\
This capsule contains code and data used for the reproducible run of the data analysis used in "ILT2 identifies an unexploited pool of intratumoral CD8+ bystander T cells with TCR-independent cytotoxicity in renal cell carcinoma" publication.\\
\\
Dzenis Koca](/content/explore/d764e9e0-bcbc-4e82-b9bf-161bf94611e2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8089715/tree/v1/index.html)

[Physics\\
\\
4 \| Jun \| 2026\\
\\
Multi-parallelized quantum long short-term memory for enhanced time-series forecasting of nonlinear dynamics\\
\\
This script implements training, resuming, and evaluation of the Multi-Parallelized Quantum Long Short-Term Memory (MP-QLSTM) model for multi-step time-series prediction.\\
The model integrates multiple variational quantum circuits (VQCs) within each LSTM gate to enhance representational capacity compared to conventional QLSTM architectures.\\
\\
Makoto Takagi et al.](/content/explore/2f03aba7-1794-4816-aab4-8eabd8cbce90?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2282885/tree/v1/index.html)

[Engineering\\
\\
3 \| Jun \| 2026\\
\\
A Wireless Bidirectional Neural Interface with Neural-Signal-Dependent Self-Fine-Tuning for Closed-Loop Motor Modulation\\
\\
This is a Python example described in the article **A Wireless Bidirectional Neural Interface with Neural-Signal-Dependent Self-Fine-Tuning for Closed-Loop Motor Modulation** which utilizes sciatic nerve electrical signals as feedback and applies spinal cord stimulation to enable a paralyzed Beagle to walk under modulation. It demonstrates how the fine-tuning unit operates throughout the entire regulation process.\\
\\
The input data include neural signals recorded from two beagle dogs during the gait recovery process, as well as the stimulation current magnitude, the start and time of stimulation and other parameters recorded by the stimulator. The _Exp0\_Previous\_Modulation_ folder contains data from the first beagle used in the original experiment, while the remaining folders contain one month of neural signal recordings from the second beagle. Based on the feedback principle outlined in the article, the fine-tuning unit predicts the new stimulation current according to the above input signals to achieve the next round of spinal cord stimulation. After regulation is completed, the system enters a stable operating state. If an unusual event occurs, causing an abnormally high neural response to a particular stimulation, the system will be reactivated, sharply reduce the stimulation amplitude, and restart the closed-loop regulation process.\\
\\
Ziyao Zhao](/content/explore/221e8df3-ad30-441f-9643-8aac03f03dae?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7784598/tree/v1/index.html)

[Biology\\
\\
3 \| Jun \| 2026\\
\\
Humanized antibody-drug conjugates for SSTR2-positive tumor treatment\\
\\
This capsule contains the R scripts and reproducible computational environment (renv-locked) used for RNA-seq and pathway analysis in the study of "Humanized antibody-drug conjugates for SSTR2-positive tumor treatment".\\
\\
Zhuoxin Zhou et al.](/content/explore/1e3d2c51-73a7-4cea-83de-5afa994c6e2b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7105524/tree/v1/index.html)

[Engineering\\
\\
2 \| Jun \| 2026\\
\\
GVG: Genetic Volterra-Subspace Generator for Behavioral Modeling and Digital Predistortion\\
\\
This capsule provides a reproducible MATLAB implementation of the Genetic Volterra-Subspace Generator (GVG), an evolutionary framework for the automated generation and optimization of Volterra-based behavioral models. The proposed approach explores the canonical Volterra subspace through genetic operators to identify compact and accurate model structures suitable for RF power amplifier modeling and digital predistortion (DPD) applications.\\
\\
The capsule includes a self-contained demonstration workflow, example datasets, and scripts required to reproduce the main results presented in the associated publication. Benchmark comparisons against conventional Volterra-derived model families are provided to illustrate the effectiveness of the proposed methodology in terms of modeling accuracy and complexity reduction.\\
\\
This implementation is intended for researchers and practitioners working on behavioral modeling, power amplifier linearization, digital predistortion, and machine-learning-assisted RF system design.\\
\\
Juan A. Becerra et al.](/content/explore/3e8ffef3-9225-4201-9668-c104225cc26e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0406486/tree/v1/index.html)

[Computer Science\\
\\
2 \| Jun \| 2026\\
\\
RIS-Kernel: A Model-Agnostic Architecture for Long-Context LLM Inference via Sparse Attention\\
\\
Processing and extracting information from long scientific documents through Large Language Models (LLMs) requires holding tens of thousands of tokens in context simultaneously. However, full self-attention scales quadratically (O(N^2)), causing memory and compute requirements to quickly exceed the capacity of commodity hardware and confining deep document analysis to institutions with large GPU clusters.\\
\\
To address this constraint, RIS-Kernel is introduced as a model-agnostic inference engine that reduces self-attention arithmetic complexity to O(NlogN) while maintaining a lightweight O(N^2) sparse boolean matrix allocation that fits comfortably within commodity RAM limits. By injecting runtime sparsification directly into unmodified pre-trained models without modifying weight parameters, RIS-Kernel enables long-context processing on unaccelerated CPU hardware.\\
\\
Under controlled evaluations at 32,768 tokens (where native dense attention is feasible as a ground-truth upper bound), RIS-Kernel at 1% density and 70 ensemble seeds achieves 75.00% accuracy, exceeding the native dense baseline (71.88%). In scalability runs at 65,536 tokens—where dense attention triggers Out-of-Memory (OOM) errors—RIS-Kernel runs stably on a CPU server, showing factual retrieval gains of up to 14.06 percentage points over the zero-context floor.\\
\\
Anderson Santos](/content/explore/ccea3965-98d7-4de8-b417-940be90b649d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0406992/tree/v1/index.html)

[Engineering\\
\\
2 \| Jun \| 2026\\
\\
Incentive Stackelberg games for mean-field control with large-scale followers\\
\\
This capsule provides the MATLAB codes used to reproduce the numerical results reported in the IJDC paper "Incentive Stackelberg games for mean-field control with large-scale followers". \\
The codes implement the simplified Newton method for solving the incentive Stackelberg game associated with large-scale stochastic mean-field systems. The capsule includes the scripts for Example 5.1 and Example 5.2, where the second example corresponds to an application to cooperative adaptive cruise control (CACC). It also includes a simulation script for comparing the trajectories generated by the full-order and approximate feedback strategies in the CACC example. The outputs include iteration residuals, solution differences, cost comparisons, and trajectory figures.\\
\\
Zihang Tian](/content/explore/f9c8b07b-ce28-4c19-88a7-f16ddd74e8e5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2421593/tree/v1/index.html)

[Bioinformatics\\
\\
2 \| Jun \| 2026\\
\\
annoreport: an interactive tool for metagenome annotation\\
\\
annoreport is a lightweight, dependency-free Python tool that addresses a practical gap in metagenome-assembled genome (MAG) annotation workflows. It accepts output from either Prokka or Bakta, automatically detects the annotation tool used, and produces an interactive HTML report and TSV summary aggregating gene product frequencies, functional clustering via UniProt enrichment, hypothetical protein rates, and feature type distributions across all bins, with no external dependencies beyond Python 3.9+.\\
\\
Kepler Ridge & Byron Adams](/content/explore/fa19f633-cfbd-4daa-8ce5-2630b99b4a8d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2024669/tree/v1/index.html)

[Earth Sciences\\
\\
1 \| Jun \| 2026\\
\\
Automatic Reconstruction of Sparse Lunar Penetrating Radar Data from Chang'e-4 Rover\\
\\
This repository provides the source code associated with the manuscript “Automatic Reconstruction of Sparse Lunar Penetrating Radar Data from Chang'e-4 Rover”. The code implements an automated reconstruction workflow for sparse valid-segment extraction and stitching from Chang'e-4 Lunar Penetrating Radar (LPR) data. The workflow includes corrupted-trace detection and correction, automatic depth truncation, Sobel-X edge enhancement, IsoData-based variance thresholding, morphology-based segment refinement, reconstruction of valid radargrams, and quantitative evaluation against manual annotations. A small example dataset and evaluation files are included to demonstrate the execution workflow. The raw Chang'e LPR data are not redistributed in this repository and should be obtained from the official data release system.\\
\\
HuangJun](/content/explore/8892679d-6d11-437b-8164-c88e7684de64?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1030257/tree/v1/index.html)

[Computer Science\\
\\
1 \| Jun \| 2026\\
\\
AdaptExit-Det: Resource-Aware Exit Decisions for Real-Time Object Detection\\
\\
This capsule reproduces the quick verification path for AdaptExit-Det, a resource-aware early-exit strategy for real-time object detection. It validates precomputed benchmark artefacts, regenerates key paper figures, and exports a verification report and metrics snapshot for submission support.\\
\\
Low Choon Keat](/content/explore/ec89182d-9beb-45a0-b7a4-125220f8a2c0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4389765/tree/v1/index.html)

[Medical Sciences\\
\\
1 \| Jun \| 2026\\
\\
Vestibular-GenBN: A modular generative Bayesian network framework for vestibular diagnostic knowledge engineering\\
\\
Vestibular-GenBN is an open-source Python seed framework for modular generative Bayesian diagnostic networks in vestibular medicine.\\
\\
Dong Woo Nam & Ja-Won Koo](/content/explore/64f90b3e-89f7-4cae-bc9c-b7f2bb4a7172?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3928062/tree/v1/index.html)

[Biology\\
\\
1 \| Jun \| 2026\\
\\
Mechanical coupling between the cytoplasmic membrane and the cell wall shapes bacterial envelope mechanics\\
\\
MATLAB code and data for regenerating figures from the manuscript “Mechanical coupling between the cytoplasmic membrane and the cell wall shapes bacterial envelope mechanics.”\\
\\
Jiawei Sun](/content/explore/615ab467-69d8-4929-9bb4-ef7ea93d9a3e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5415414/tree/v1/index.html)

[Computer Science\\
\\
1 \| Jun \| 2026\\
\\
FUGCF: Training-free and Unbiased Graph Collaborative Filtering for Personalized Recommendations\\
\\
Code and Data for the paper: Training-free and Unbiased Graph Collaborative Filtering for Personalized Recommendations\\
\\
Ziyang Liu et al.](/content/explore/5f13e6b6-0e07-4fe7-8c23-273738e9259d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7145941/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tkde.2026.3669816) published in [IEEE Transactions on Knowledge and Data Engineering](/content/explore?query=IEEE%20Transactions%20on%20Knowledge%20and%20Data%20Engineering&refine=journal/index.html)

[Computer Science\\
\\
1 \| Jun \| 2026\\
\\
Securing Metaverse Blockchain Infrastructure: LLM-Assisted Vulnerability Detection on Solana and Algorand Smart Contracts](/content/explore/1f058ad9-9276-4b65-8559-c1fbd4d6c064?page=2&filter=all/index.html)

[The overwhelming majority of LLM-based smart\\
contract audit tools target Ethereum and the EVM toolchain,\\
leaving Solana and Algorand, the second and third largest\\
blockchain ecosystems by developer activity, essentially without\\
automated vulnerability detection support. This paper presents\\
the first systematic benchmark of frontier large language mod-\\
els on non-EVM smart contract vulnerability detection. We\\
construct a dataset of 24 contract instances spanning eight\\
vulnerability classes across two chains: five Solana/Anchor classes\\
(missing signer check, account confusion, arithmetic overflow,\\
bump seed canonicalization, and stale cross-program invocation\\
data) and three Algorand/PyTEAL classes (logic signature abuse,\\
group transaction manipulation, and unchecked asset fields).\\
Each instance has a paired patched version to enable false\\
positive rate measurement. We evaluate three frontier models\\
(GPT-4o, Claude Sonnet 4, and Llama-3.3-70B-Instruct) under\\
three prompting strategies: zero-shot, chain-of-thought (CoT),\\
and retrieval-augmented generation (RAG). Across 216 experi-\\
mental runs we measure detection rate (DR), false positive rate\\
(FPR), explanation quality score (EQS on a 1 to 5 rubric), and\\
reasoning coherence (RC), defined as the absence of EVM-specific\\
hallucinations in non-EVM contexts. Our primary findings are:\\
(1) CoT prompting achieves 100% detection rate and 0% false\\
positive rate across all three models; (2) overall FPR is 5.1%\\
(below the 0.10 threshold), though three of the nine strategy-\\
model pairs exceed it, with RAG + Claude Sonnet 4 highest\\
at 16.7%; (3) zero-shot performance drops to 33% DR on the\\
most subtle Solana class (V2 account confusion); (4) Claude\\
Sonnet 4 produces the highest explanation quality (mean EQS\\
4.24 vs. 2.88 to 3.01 for the other two models); and (5) EVM-\\
specific hallucinations occur in up to 25% of Llama-3.3-70B\\
zero-shot Algorand runs, confirming PyTEAL’s limited repre-\\
sentation in open-model pre-training corpora. We release all\\
contracts, prompts, raw model outputs, and scoring code at](/content/explore/1f058ad9-9276-4b65-8559-c1fbd4d6c064?page=2&filter=all/index.html) [https://github.com/NucleiAv/llm-audit-nonevm](https://github.com/NucleiAv/llm-audit-nonevm).

Anmol Vats

[Open Capsule](/content/capsule/7940624/tree/v1/index.html)

[Medical Sciences\\
\\
28 \| May \| 2026\\
\\
Vestibular-GenBN: An open-source framework for modular generative Bayesian networks in vestibular diagnostic knowledge engineering\\
\\
Vestibular-GenBN is an open-source Python seed framework for modular generative Bayesian diagnostic networks in vestibular medicine.\\
\\
Dong Woo Nam & Ja-Won Koo](/content/explore/88404082-317b-484e-aa7b-ad0714db79df?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9588443/tree/v1/index.html)

[Earth Sciences\\
\\
28 \| May \| 2026\\
\\
Slow slip modulates low-frequency seismicity on the San Andreas Fault\\
\\
This capsule provides a three-step pipeline for detecting and clustering strain transients in borehole strainmeter data using wavelet transforms and deep learning (AutoencoderZ). The workflow is designed for the automated detection of Slow Slip Events (SSEs).\\
\\
Zahra Zali](/content/explore/704b953c-caa4-4c0e-aba3-3add8f301587?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0214851/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
27 \| May \| 2026\\
\\
Computing Scaled Relative Graphs of Discrete-time LTI Systems from Data\\
\\
Code to Compute SRGs of discrete time systems from state-space representations or data.\\
\\
Talitha Nauta](/content/explore/2fe6a647-e906-496b-8518-61ed5c18748d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0781271/tree/v1/index.html)

[Computer Science\\
\\
26 \| May \| 2026\\
\\
Leader-Follower Optimizer (LFO)\\
\\
Implementation of the Leader–Follower Optimizer (LFO), a population-based metaheuristic with explicit leader-follower dynamics and hierarchical influence mechanisms.\\
\\
Bruno Luiz Pereira](/content/explore/79b71a4e-75e3-4954-b33a-74042ef613d1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7665620/tree/v1/index.html)

[Computer Science\\
\\
26 \| May \| 2026\\
\\
Diffusion-Regularised Autoencoder\\
\\
This capsule contains the code for a diffusion-regularised autoencoder trained on CIFAR-10. The model combines a UNet-style convolutional autoencoder with a latent-space diffusion prior (SmallTransformerDenoiser) for high-fidelity image reconstruction. The pipeline includes autoencoder warm-start training, joint optimisation of the decoder and denoiser with the encoder frozen, and full evaluation reporting PSNR, SSIM, LPIPS, and FID metrics. Ablation studies over latent dimensions and diffusion schedules are also included.\\
\\
ABSTRACT: High-fidelity image reconstruction using autoencoders is often limited by a trade-off between\\
pixel-level accuracy and the quality of latent representations. Conventional autoencoders can achieve low reconstruction error but often produce poorly structured latent spaces, whereas diffusion-based generative models offer strong perceptual quality at the cost of high computational complexity and iterative sampling. This paper proposes a diffusion-regularised autoencoder that integrates latent-space diffusion as a trainingtime regularisation mechanism rather than a generative inference process. The framework consists of a UNet-style convolutional autoencoder coupled with a lightweight Transformer-based denoiser operating on the latent representation. Training is performed in two stages: an initial autoencoder warm-start optimised for reconstruction fidelity, followed by joint optimisation of the decoder and denoiser with the encoder frozen. A dynamically scaled composite loss balances reconstruction accuracy with diffusion regularisation,\\
while a cosine noise schedule and exponential moving average parameter tracking are employed to stabilise optimisation. Experiments on the CIFAR-10 dataset demonstrate that the proposed method achieves nearlossless reconstruction performance, with a Peak Signal-to-Noise Ratio (PSNR) of 45.15 dB, a Structural Similarity Index Measure (SSIM) of 0.9982, a Learned Perceptual Image Patch Similarity (LPIPS) of 0.0017, and a reconstruction Fréchet Inception Distance (rFID) of 0.736. Qualitative evaluations further show smooth latent interpolations and well-organised latent manifolds. The results indicate that diffusionbased regularisation offers an effective and computationally efficient means of enhancing autoencoder representations for high-fidelity image reconstruction.\\
\\
Onyebuchi Vincent Edigbo & Shivang Shukla](/content/explore/189e217a-e82d-4c4d-9f8e-60285f64b60b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8296884/tree/v1/index.html)

[Computer Science\\
\\
26 \| May \| 2026\\
\\
JDCNet: Confidence-Gated Privileged-Modality Distillation for Cost-Preserving X-ray Inference\\
\\
JDCNET Source Code\\
\\
Bo Ma](/content/explore/0a9d05ef-507e-4ae8-bfd2-5bbb636ee5bd?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4742424/tree/v1/index.html)

[Computer Science\\
\\
26 \| May \| 2026\\
\\
COREY: Entropy-Guided Runtime Chunk Scheduling for Selective Scan Kernels\\
\\
model quantization\\
\\
Bo Ma](/content/explore/f5db24bd-358c-45ee-9b9f-12d315e22c07?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3457679/tree/v1/index.html)

[Engineering\\
\\
26 \| May \| 2026\\
\\
A Structured Family of Grassmannian Constellations via Geodesic Mapping for MIMO Noncoherent Communications\\
\\
Structure family of Grassmannian constellations for MIMO noncoherent communications.\\
\\
Álvaro Pendás-Recondo & Enrique Pendás-Recondo](/content/explore/7705ab27-ce6f-468f-b216-81bace408578?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2319310/tree/v1/index.html)

[Engineering\\
\\
26 \| May \| 2026\\
\\
Online Diagnosis for Transformer Winding Defect via CACP and CCO-LCE Algorithm\\
\\
**Online Diagnosis for Transformer Winding Defect via CACP and CCO-LCE Algorithm** \\
\\
This capsule contains the reproducible code for the two core modules requested by the reviewers of the paper **Online Diagnosis for Transformer Winding Defect via CACP and CCO-LCE Algorithm** submitted to IEEE Transactions on Industrial Electronics (TIE).\\
\\
**Overview** \\
\\
Power transformers are critical equipment in power systems, and winding deformation caused by short-circuit impacts is a major cause of transformer failures. Frequency Response Analysis (FRA) is the standard method for winding condition assessment, but its interpretation relies heavily on expert experience. This work proposes an intelligent diagnosis framework that combines deep learning and ensemble learning to achieve automatic and high-precision identification of winding fault types, locations and severities.\\
\\
**Code Structure**\\
\\
- `/ACP/`: Attention Clustering Pooling module implementation\\
  - `__init__.py`: Module initialization file\\
  - `acp.py`: ACP module in the CACP feature extraction method\\
- `/CCO/`: CCO-optimized LCE classifier implementation\\
  - `__init__.py`: Module initialization file\\
  - `cco_code.py`: Core implementation of the CCO optimization algorithm\\
  - `ben_functions.py`: Benchmark function definitions\\
  - `fun_range.py`: Parameter range definitions\\
  - `example_usage.py`: Usage example\\
  - `main.py`: Main entry script for the complete diagnosis pipeline\\
\\
XiaoboZhang](/content/explore/f79ac566-5ba2-4892-8430-e80f4490b5eb?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0749191/tree/v1/index.html)

[Engineering\\
\\
26 \| May \| 2026\\
\\
Image data reconstruction-driven AI framework enables robust and high-fidelity defect diagnosis in solution-processed manufacturing\\
\\
- Code for Grid-Based Data Boosting (GDB)\\
- Code for AI model design\\
\\
Sivaranjini Mohanan, Nagesh Pandey & Jongsu Lee](/content/explore/989ed33a-4db6-402d-8bf6-939ae721c409?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8211557/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
Speaker-Disentangled Chunk-Wise Regression for Syllabic Tokenization\\
\\
Reproduction of the main results (Table 1) and downstream performance (Tables 4 and 5) in the paper "Speaker-Disentangled Chunk-Wise Regression for Syllabic Tokenization"\\
\\
Note:\\
Results for SylReg-LM-7B as well as perplexity and auto-BLEU metrics in Table 4 are not available due to out of memory with codeocean's computational resources\\
\\
Ryota Komatsu, Kota Kawakita, Takuma Okamoto & Takahiro Shinozaki](/content/explore/cd446fe6-a9dc-4a5b-ad0f-7d0c101be5e4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6620861/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
Evaluating Hybrid Automata Learning Tools Based on their Success in Verification\\
\\
This capsule implements a novel accuracy metric for evaluating the performance of hybrid automata learning approaches and tools. We demonstrate this new metric on several hybrid system benchmarks, including an industrial case study featuring a boost-converter circut. The evaluation provided by this capusule compares three state-of-the-art automaton learning\\
approaches that exhibit distinct methodological characteristics.\\
\\
Niklas Kochdumper et al.](/content/explore/62ad5cb2-bb29-475b-bb7f-b446cfbd65ed?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5093493/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
REAEDP: Entropy-Calibrated Differentially Private Data Release with Formal Guarantees and Attack-Based Evaluation\\
\\
REAEDP Source Code\\
\\
Bo Ma](/content/explore/0c283027-2e60-41e0-9569-f9b7528f1e07?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6034180/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
PPEDCRF: Privacy-Preserving Enhanced Dynamic CRF for Location-Privacy Protection for Sequence Videos with Minimal Detection Degradation\\
\\
PPEDCRF Source Code\\
\\
Bo Ma](/content/explore/486a9131-28cf-4e52-ab2b-6201aacb019e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9207610/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
Bodhi VLM: Privacy-Alignment Modeling for Hierarchical Visual Representations in Vision Backbones and VLM Encoders via Bottom-Up and Top-Down Feature Search\\
\\
Bodhi VLM source code\\
\\
Bo Ma](/content/explore/450f58b4-a40c-4360-ae73-30eb3eecda10?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8370806/tree/v1/index.html)

[Bioinformatics\\
\\
25 \| May \| 2026\\
\\
Vestibular-BayesSeed: An open-source framework for evidence-anchored logistic Bayesian networks in dizziness diagnosis\\
\\
**Vestibular-BayesSeed** is an open-source seed framework for constructing evidence-anchored logistic Bayesian diagnostic networks for vestibular disorders.\\
\\
Dong Woo Nam](/content/explore/ae1e7727-de06-4ec1-88a4-8636f89657a4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6831105/tree/v1/index.html)

[Engineering\\
\\
25 \| May \| 2026\\
\\
Theoretical Modeling of Ising Spin Communication Channels with Relaxed-State Modulation\\
\\
This package reproduces the main building blocks of the relaxed-state Ising communication channel from~\[1\] in a small, self-contained form. Running main.py samples random Ising media, writes a classical bit by locally perturbing the couplings and fields near the transmit sites, relaxes each medium to its ground state, and then decodes at Bob (a private differential readout) and at Charlie (a passive eavesdropper that sees only public features). It harvests channels that give Bob a clear advantage over Charlie and recomputes their bit-error rates (BERs) with more symbols. Everything --- media, endpoints, write perturbations, transmitted bits, and dynamic noise --- is generated at run time, so no external data files are needed.\\
\\
Burhan Gulbahar](/content/explore/aa052ef0-6b69-45de-afea-9c3017853ad6?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8532676/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
Dynamically Gated TinyMLPs for Extreme Edge Predictive Maintenance: A Noise-Robust Feasibility Study\\
\\
Abstract—The rapid proliferation of Industrial Internet of Things (IIoT) micro-nodes has accelerated the demand for resource-constrained, edge-deployable predictive maintenance (PdM) solutions. However, deploying standard Deep Neural Networks on hardware-constrained edge nodes is severely limited by memory footprint, computational latency, and vulnerability to industrial sensory noise. \\
\\
To address these challenges, this study presents a focused feasibility analysis of a highly compact, trainable input-filtering layer—termed the NoiseGate—integrated into a hardware-constrained TinyMLP (16-8 hidden dense layers). The custom element-wise NoiseGate layer dynamically filters high-frequency sensor noise before classification by learning feature-specific gating weights without adding significant parameters or computational overhead. We evaluate the proposed NoiseGate TinyMLP against standard Logistic Regression, standard TinyMLP, and Gaussian Noise-regularized networks across two prominent benchmarks: the AI4I 2020 Predictive Maintenance Dataset and the MetroPT Air Compressor Dataset. \\
\\
Our findings demonstrate that the NoiseGate TinyMLP establishes a superior precision-recall balance under varying noise conditions, achieving a peak PR-AUC of 0.9934 on the MetroPT benchmark. The results prove that dynamically gating sensory features inside extremely compact networks can successfully reconcile the conflicting requirements of extreme edge memory limitations and robust classification reliability under environmental perturbations.\\
\\
Aman Sharma](/content/explore/cb92a18d-3d2a-4787-a883-1df5834f6479?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7559748/tree/v1/index.html)

[Computer Science\\
\\
25 \| May \| 2026\\
\\
A Unified Engineering-Informatics Framework for Noise-Robust and Explainable Edge Predictive Maintenance\\
\\
Abstract—Modern industrial systems generate high-velocity, high-dimensional sensory telemetry, requiring robust anomaly detection and predictive maintenance (PdM) models. While edge-computing architectures mitigate bandwidth requirements by performing inference locally, resource-constrained edge devices are highly vulnerable to noisy operational conditions, severe class imbalance, and lack of transparency. \\
\\
To resolve this multi-dimensional challenge, we introduce a unified, reliability-aware engineering-informatics framework for compact edge predictive maintenance. The framework integrates three core innovations: (1) a custom, magnitude-aware PhysicsGate layer that dynamically weights incoming features based on physical operational scale, (2) a multi-objective Joint Autoencoder-Classifier model that leverages unsupervised reconstruction loss to regularize class-imbalanced failure boundaries, and (3) a path-based Explainable AI (XAI) routine utilizing Integrated Gradients to mathematically verify feature-importance profiles. \\
\\
We audit a diverse registry of nine sequential, residual, and gated model families (including ResMLP-Lite and GRU-Lite) across 1,516,948 operational telemetry rows from the MetroPT compressor and AI4I benchmarks. The results demonstrate that recurrent baselines achieve competitive discriminative capabilities (peak PR-AUC of 0.9950), while physics-gated joint architectures exhibit superior probability calibration (minimized Brier scores) and exceptional noise resilience under simulated sensor dropout and Gaussian perturbations. The Integrated Gradients attribution profiles confirm that our gated models successfully align with expected thermodynamic and mechanical operational boundaries, establishing a new benchmark for reliable, explainable, and ultra-compact edge industrial intelligence.\\
\\
Aman Sharma](/content/explore/e02c9ed4-e93c-4368-94e6-fc4add0c72e0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8855575/tree/v1/index.html)

[Social Sciences\\
\\
25 \| May \| 2026\\
\\
Conceptual priorities shape individual gaze patterns during naturalistic visual attention\\
\\
Code and data to reproduce results reported in Haskins, Packard, & Robertson 2026\\
\\
Amanda J. Haskins, Katherine O. Packard & Caroline E. Robertson](/content/explore/f9f0d56c-8fa0-4d3f-a066-47c31808313f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2103341/tree/v1/index.html)

[Computer Science\\
\\
22 \| May \| 2026\\
\\
"CICBLS-IRR: A Class-Incremental Cascade Broad Learning System via Incremental Ridge Regression for Catastrophic Forgetting Mitigation"\\
\\
This compute capsule contains the official MATLAB implementation for the proposed CICBLS-IRR (Class-Incremental Cascade Broad Learning System via Incremental Ridge Regression). Designed to tackle the catastrophic forgetting problem in class-incremental learning (CIL), CICBLS-IRR dynamically expands its cascade architecture and updates weights without requiring extensive retraining of historical data.\\
\\
YuJie Fang ChangZhou University](/content/explore/a2a229b6-b6fd-4628-8e63-74c4c79f826c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9293100/tree/v1/index.html)

[Computer Science\\
\\
22 \| May \| 2026\\
\\
A Post-Quantum Motivated Resultant-Based Proof-Of-Work With Lightweight Verification\\
\\
This capsule contains the prototype implementation \\
of SGPoW (Small-Generator Proof-of-Work), a \\
post-quantum-motivated resultant-based proof-of-work \\
defined over cyclotomic rings. The code reproduces \\
the benchmark results reported in the IEEE Access \\
manuscript "A Post-Quantum-Motivated Resultant-Based \\
Proof-of-Work with Lightweight Verification".\\
\\
Ebru Adiguzel Goktas & Bayram Ali Ersoy](/content/explore/c631bd0e-709a-4695-ab1e-e6adc346dcd4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5471230/tree/v1/index.html)

[Physics\\
\\
22 \| May \| 2026\\
\\
Overcoming the performance ceiling of textured piezoelectric ceramics\\
\\
Code for Overcoming the performance ceiling of textured piezoelectric ceramics.\\
\\
Jinjing Zhang et al.](/content/explore/dd8243e0-21f1-4e96-b73e-6532273f8e3f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1495815/tree/v1/index.html)

[Physics\\
\\
22 \| May \| 2026\\
\\
GRAPE: A symplectic integrator for a native relativistic orbitography software\\
\\
GRAPE (General Relativity Accelerometer-based Propagation Environment) is a Julia-based framework for simulating spacecraft trajectories in a fully relativistic formulation. It integrates the motion of a spacecraft within arbitrary spacetime metrics (Schwarzschild, Kerr, Newtonian approximations, etc.), including non-gravitational forces and accelerometer-based models.\\
\\
Jean Pierre BARRIOT](/content/explore/ee3c3d90-c695-48bb-a6f8-1cd426f4c24e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4754289/tree/v1/index.html)

[Computer Science\\
\\
22 \| May \| 2026\\
\\
FusionXNet-Edge: A Physics-Guided Lightweight Framework for Predictive Maintenance Under Cross-Domain Deployment Shift\\
\\
Academic reproducibility capsule containing the FusionXNet-Edge source package, validation scripts, archival notebooks, saved lightweight model artifacts, dataset references, and reference results for predictive maintenance under cross-domain deployment shift.\\
\\
Aman Sharma, Kwan Yong Sim & Sivachandran Chandrasekaran](/content/explore/9098ad7d-a259-4fc3-80d2-97f1d56a6186?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9506354/tree/v1/index.html)

[Computer Science\\
\\
22 \| May \| 2026\\
\\
Supplement for: Soft-VAP: A Learned Decoder for Frozen Voice Activity Projection States\\
\\
Voice Activity Projection (VAP) predicts a distribution over 256 future two‑speaker activity states from stereo audio. The standard fixed decoder maps all 175 simultaneous‑activity states to one overlap label, so its backchannel (BC) recall is zero on overlap windows. Soft‑VAP replaces only this output map with a 514‑parameter logistic‑regression decoder and leaves the VAP backbone frozen. On conversation‑disjoint Switchboard splits, validation selects a 0.2 s pooling window. The decoder reaches 0.583 BC‑vs‑interruption (BC‑vs‑I) macro F1 on the balanced internal test, compared with 0.333 for the fixed mask and 0.368 for a duration rule. A conversation‑level cluster bootstrap estimates a 0.215 macro‑F1 gain over the duration rule (95% confidence interval (CI): 0.192–0.236). A 6000‑window natural‑prior audit gives the same ranking (0.535, 0.413, and 0.251). Reduced‑feature probes show that pairwise bit features account for much of the gap between 8‑bit marginals (0.497) and the full 256‑state vector (0.583).\\
\\
Ying Xu](/content/explore/1ecfe997-5d54-4978-bd47-464254102934?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1921184/tree/v1/index.html)

[Mathematics\\
\\
20 \| May \| 2026\\
\\
MooAFEM: Numerical investigation of an iterative Galerkin method driven by an elliptic reconstruction estimator for quasilinear elliptic PDEs\\
\\
We investigate an iterative Galerkin method for quasilinear elliptic problems in the Browder-Minty setting. The resulting discrete nonlinear systems are solved by linearization via a (damped) Zarantonello iteration. Here, adaptive mesh refinement is driven by an elliptic reconstruction error estimator, which is natural in the sense that the a posteriori bounds for the linearization and discretization errors are well separated. For this setting, the algorithm ensures unconditional full R-linear convergence and, for sufficiently small adaptivity parameters, optimal convergence rates with respect to the overall computational cost. The numerical experiments reproduce the results reported in the associated publication. They focus on the comparison between the reconstruction and standard estimators and the influence of the chosen scalar product in the Zarantonello iteration.\\
\\
Maximilian Brunner, Gregor Gantner, Christoph Lietz & Dirk Praetorius](/content/explore/9a29d4df-7bb6-47ae-ab56-1b9ec17c8cf6?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8818408/tree/v1/index.html)

[Medical Sciences\\
\\
19 \| May \| 2026\\
\\
Detangling the spinal respiratory network’s responses to cervical epidural stimulation after spinal cord injury (VGLUT VIAAT)\\
\\
Analysis of FOS coexpression with VIAAT and VGLUT in the cervical spinal cord after injury and electrical stimulation.\\
\\
Alyssa Mickle, Jesús Peñaloza-Aponte, Caitlin Brennan & Erica Dale](/content/explore/0c7cff8f-54a4-4d69-87b6-c34adfc7ca98?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9497078/tree/v1/index.html)

[Medical Sciences\\
\\
19 \| May \| 2026\\
\\
Detangling the spinal respiratory network’s responses to cervical epidural stimulation after spinal cord injury (SST CHAT)\\
\\
Co-expression analysis of FOS with CHAT and SST in the cervical spinal cord.\\
\\
Alyssa Mickle, Jesús Peñaloza-Aponte, Caitlin Brennan & Erica Dale](/content/explore/105365e2-d644-4996-b860-edf1c1b4c11a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2409387/tree/v1/index.html)

[Computer Science\\
\\
19 \| May \| 2026\\
\\
Zero-Shot Urban Microclimate Estimation Through Image-Based Cross-Attention Learning\\
\\
This capsule enables replication of the results in the paper titled "No Sensors? No problem: Zero-Shot Urban Microclimate Estimation Through Image-Based Cross-Attention Learning," currently under revision at IEEE Access.\\
\\
Wataru Kunimi et al.](/content/explore/e1f0897d-20e3-4961-b53c-f80a4ef7dcbe?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7863392/tree/v1/index.html)

[Physics\\
\\
19 \| May \| 2026\\
\\
χ as a Transition Corridor Rather Than a Relaxation Attractor: Emergent χ-Corridor Formation in Recursive Coupling-Damping Systems\\
\\
**Code Ocean Description**\\
\\
text\\
\\
Copy\\
\\
```text\\
This capsule investigates the emergence of a persistent χ-near stability corridor in recursive coupling-damping systems under χ-blind asymmetric evolution.\\
\\
The irrational stability ratio:\\
\\
χ ≈ 0.5512855984\\
\\
is intentionally excluded from the governing evolution equations and used only afterward as a measurement reference.\\
\\
A fine asymmetry sweep across seven independent random seeds demonstrates the emergence of a stable χ corridor extending from asymmetry strength 5.82 through 8.20, with best mean-ratio agreement reaching:\\
\\
C/K = 0.5513233716\\
\\
with absolute distance from χ:\\
\\
3.78 × 10^-5\\
\\
while maintaining complete system survival.\\
\\
The results support the interpretation of χ as a stable transition corridor within a recursive branch manifold rather than a conventional relaxation attractor.\\
\\
Included:\\
- simulation scripts\\
- fine sweep datasets\\
- aggregate outputs\\
- phase maps\\
- corridor occupancy analysis\\
- survival/failure analysis\\
- reproducible multi-seed simulations\\
```\\
\\
Matthew J. Hall (0009-0001-7066-2558)](/content/explore/5ff4ab20-6a34-46ad-89ed-4e6d4fd5288f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0677887/tree/v1/index.html)

[Physics\\
\\
19 \| May \| 2026\\
\\
Recursive Sphere Stabilization and Dimensional Shell Formation: An 11D Chronos Stability Framework](/content/explore/50a51f73-4807-4b0c-8f11-a97ffa242907?page=2&filter=all/index.html)

[This capsule provides a reproducible computational implementation of the recursive dimensional hierarchy proposed in:\\
\\
Recursive Sphere Stabilization and Dimensional Shell Formation: An 11D Chronos Stability Framework (V2.2)\\
\\
The simulation operationalizes the Chronos recursive stability framework through a sequence of dimensional emergence stages beginning from a distributed recursive stability field (D0) and progressing through localized condensation, recursive spiral motion, cone expansion, toroidal closure, spherical shell stabilization, linked sphere chains, compression-expansion dynamics, higher toroidal circulation, stable radius distributions, and a full recursive stability envelope (D11).\\
\\
The framework is governed by the Chronos stability condition:\\
\\
chi\_eff = C / K -> chi\\
\\
where stable recursive systems persist only within bounded regions between collapse and blow-up regimes.\\
\\
This capsule generates:\\
\\
- Recursive geometric visualizations,\\
- Stability-envelope simulations,\\
- Linked sphere-chain coupling data,\\
- Radius-distribution filtering near chi,\\
- Compression-expansion dynamics,\\
- And operational CSV outputs for further analysis.\\
\\
The purpose of this capsule is not to claim final empirical validation, but to provide an operational and reproducible computational demonstration of the recursive dimensional framework described in the accompanying paper.](/content/explore/50a51f73-4807-4b0c-8f11-a97ffa242907?page=2&filter=all/index.html)

[Associated Paper:\\
Hall, M. (2026). Recursive Sphere Stabilization and Dimensional Shell Formation: An 11D Chronos Stability Framework (V2.2). Zenodo.](/content/explore/50a51f73-4807-4b0c-8f11-a97ffa242907?page=2&filter=all/index.html) [https://doi.org/10.5281/zenodo.20261488](https://doi.org/10.5281/zenodo.20261488)

Matthew J. Hall (0009-0001-7066-2558)

[Open Capsule](/content/capsule/9857988/tree/v1/index.html)

[Engineering\\
\\
18 \| May \| 2026\\
\\
Mechanism Design and Co-operative Game Theory for Cybersecurity Defenders - Two Combined Games\\
\\
A Game Theoretic plugin for Security Engineers, to help developers write secure code. The plugin extends functionality of a security code review tool called Bandit. See this plugin's readme for more details.\\
\\
v1.0 was the main release\\
The next version implemented error handling for a corner case involving testing of 1 test file (hello.py)\\
\\
Mithun Vaidhyanathan, Weisheng Si, Bahman Javadi & Seyit Camtepe](/content/explore/0b6b779f-c591-47ef-a28e-dde771ef550b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3133433/tree/v1/index.html)

[Computer Science\\
\\
19 \| May \| 2026\\
\\
FusionNet: Intelligent Sequence Fusion for Predictive Maintenance in Edge-Enabled Industrial IoT Systems\\
\\
Source code and reproducibility capsule for FusionNet: Intelligent Sequence Fusion for Predictive Maintenance in Edge-Enabled Industrial IoT Systems. Raw datasets are not redistributed; users should obtain the cited public datasets and place them under the documented data/raw/ structure before running full experiments.\\
\\
Aman Sharma, Kwan Yong Sim & Sivachandran Chandrasekaran](/content/explore/d8702aec-19e0-4d5a-b4e1-688b371ac950?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1952235/tree/v2/index.html)

[Engineering\\
\\
18 \| May \| 2026\\
\\
Co-adaptation of dynamic human-machine interfaces\\
\\
Co-adaptive human machine interaction game\\
\\
Amber H.Y. Chou, Momona Yamagami & Samuel A. Burden](/content/explore/57670704-a36b-4239-add6-bcf8e35c33f1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5038881/tree/v1/index.html)

[Social Sciences\\
\\
15 \| May \| 2026\\
\\
A lack of clean drinking water is associated with lacking food and experiencing food safety threats in 121 countries around the world (Bruine de Bruin et al., 2026, Nature Food)\\
\\
Analysis for:\\
A lack of clean drinking water is associated with lacking food and experiencing food safety threats in 121 countries around the world (Bruine de Bruin et al., 2026, Nature Food)\\
\\
Coder: Joshua Inwald\\
\\
Wändi Bruine de Bruin](/content/explore/1a136e8a-e440-4ff2-877d-547110ae7b5e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3707067/tree/v1/index.html)

[Bioinformatics\\
\\
14 \| May \| 2026\\
\\
Examples of analytical visualizations commonly used in Nanopore Direct RNA sequencing (DRS) studies\\
\\
Examples of analytical visualizations commonly used in Nanopore Direct RNA seqiencing (DRS) studies. The code and data will continue to be updated in the future, which has helped researchers analyze DRS research data\\
\\
Tianyuan Zhang](/content/explore/ee9d7d87-6fd4-4c73-b1c9-a57cec76bb53?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8214594/tree/v1/index.html)

[Engineering\\
\\
14 \| May \| 2026\\
\\
Learning Radio Maps via Graph Transformer for User-Centric Cell-Free Massive MIMO\\
\\
The source code for paper "Learning Radio Maps via Graph Transformer for User-Centric Cell-Free Massive MIMO" submited to IEEE Transactions on Wireless Communications.\\
\\
Bin Yang](/content/explore/ede3fe7b-04c5-479d-953c-70e1983ed72c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3616726/tree/v1/index.html)

[Computer Science\\
\\
13 \| May \| 2026\\
\\
Supplement for: Densing Law Revisited for Chinese Large Language Models\\
\\
Capability density, the capability obtained per model parameter, offers a way to study efficiency in large language models (LLMs). The Densing Law reports that the open-source LLM density frontier doubles roughly every 3.5 months on English benchmarks. Chinese LLMs provide a useful test of this pattern because the language, tokenization problem, benchmark suite, and model ecosystem differ from the English setting. This study compiles benchmark records for more than 30 openweight models released from March 2023 to April 2026. It examines CEval, CMMLU, and SuperCLUE scores through parameter density, bootstrap uncertainty, cross-lingual density ratios, tokenizer-adjusted density, and inference-cost density. The analysis gives a doubling-period point estimate near 3.2 months, shows a narrowing Chinese-English density gap, and finds faster apparent densing in reasoning-heavy categories. The study contributes a reproducible design for measuring Chinese LLM efficiency and treats tokenizer measurements and estimates of floating point operations (FLOPs) as fixed-sample sensitivity measurements. The most distinctive signal is not a faster global slope, but the structure of Chinese density growth across language, task, tokenizer, and deployment dimensions.\\
\\
Anonymous](/content/explore/de9c5502-ce59-49f2-9161-b657d5ab88b8?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6505220/tree/v1/index.html)

[Computer Science\\
\\
13 \| May \| 2026\\
\\
Supplement for: When Stories Drift: Affective Context Drift in Large Language Model Narrative Generation Under Incremental Priming\\
\\
Whether evaluative context before a neutral story stem can shift generated-continuation sentiment is open for controllability and safety. We study Chinese narrative generation with four large language models (LLMs), ten stems, matched positive and negative primers, unrelated affective primers, and baselines. Across 4,000 quality-gated generations, matched primers shifted sentiment in all 24 model valence level cells after Bonferroni correction. Five of 12 unrelated-primer cells also remained significant. This pattern supports semantic prompt sensitivity and reduced-affinity affective leakage, not a strong semantic-invariant anchoring account. Leave-one-stem analysis, dual-judge overlap, a 32-item agreement check, ablation, mitigation, and a small probe bound the claim. Supplementary materials and an anonymized analysis capsule support reproduction.\\
\\
Anonymous Authors](/content/explore/88276eca-83f3-4fa5-9798-8c2600102f5b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4951744/tree/v1/index.html)

[Medical Sciences\\
\\
13 \| May \| 2026\\
\\
Self-perceived gender conformity and depressive symptoms across European gender regimes\\
\\
R-code for peer review\\
\\
Hanna Wierenga, Birgit Derntl & Pia Schober](/content/explore/4e85e8ed-01dc-46a4-8029-f168476f9e36?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7365355/tree/v1/index.html)

published in [Nature Mental Health](/content/explore?query=Nature%20Mental%20Health&refine=journal/index.html)

[Computer Science\\
\\
12 \| May \| 2026\\
\\
Optimal Deployment of Third-Party Cybersecurity Tools in Operational Technology Environments: a Power Plant Use-Case\\
\\
This Code Ocean capsule contains the code and computational resources associated with the IEEE submission "Optimal Deployment of Third-Party Cybersecurity Tools in Operational Technology Environments: a Power Plant Use-Case". The work focuses on the optimal deployment of third-party cybersecurity tools in operational technology environments, considering a power plant use-case.\\
\\
Giovanni Gaggero](/content/explore/a54ecd0f-7690-409c-bc33-016eb976dbb8?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3328627/tree/v1/index.html)

[Computer Science\\
\\
12 \| May \| 2026\\
\\
A CBOM-Based Framework for Post-Quantum Cryptographic Risk Scoring and Remediation Prioritisation\\
\\
Implementation accompanying the paper "A CBOM-Based Framework for Post-Quantum\\
Cryptographic Risk Scoring and Remediation Prioritisation". Provides a five-factor\\
weighted risk-scoring model for cryptographic assets, AHP-validated expert weights,\\
a 30-asset synthetic TLS inventory benchmark, and a 21-sector real-world TLS scan\\
of 1,050 HTTPS endpoints. Requires no third-party Python packages beyond matplotlib\\
and numpy for figure generation.\\
\\
Praveen Kumar Palaniswamy & Muthukumar Kubendran](/content/explore/ced0909e-30f6-4a1c-b725-3247c91e877b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9773956/tree/v1/index.html)

[Computer Science\\
\\
11 \| May \| 2026\\
\\
FSP Matrix Compression Distillation Algorithm Based on Krylov Subspace Projection\\
\\
Specific Implementation Based on Krylov Subspace Projection Compression Algorithm\\
\\
XuYan](/content/explore/cfd35b05-36db-45e6-81f8-521afac1653b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2219239/tree/v1/index.html)

[Computer Science\\
\\
20 \| May \| 2026\\
\\
bulletin-fetcher\\
\\
Example scripts and notebooks to test bulletin-fetcher library\\
\\
Diego González Suárez](/content/explore/32ae6b11-cb50-4b06-bfc3-84be51927ac7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0018646/tree/v2/index.html)

[Computer Science\\
\\
11 \| May \| 2026\\
\\
Energy-Efficient Temporal Adaptive Threshold Method for High-Frame-Rate Maritime Foreground Detection Using Edge CPUs\\
\\
Minimal reproducible implementation of Temporal Adaptive Threshold Foreground Detection (T-ATFD) for edge-CPU maritime foreground detection. The capsule includes a deterministic synthetic smoke test, an author-curated public 200-frame demo subset, and a minimal K-parameter sanity sweep.\\
\\
Yi-Tung Chan](/content/explore/dead7350-242c-4079-9e95-d551b5fbf70a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1023007/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/les.2026.3690721) published in [IEEE Embedded Systems Letters](/content/explore?query=IEEE%20Embedded%20Systems%20Letters&refine=journal/index.html)

[Biology\\
\\
8 \| May \| 2026\\
\\
Comparing T-cell densities across histologic types in preinvasive colorectal lesions\\
\\
This capsule provides code for the T-cell density analysis in the article "Differences in T-cell densities and neighborhood patterns in human colorectal adenomas and sessile serrated lesions". Densities for CD4+, CD8+, CD4+TBX21+, CD4+RORC+, CD4+FOXP3+, CD8+TBX21+, CD8+RORC+, and CD8+FOXP3+ T-cell types are evaluated by histologic type: tubular adenomas (TA), tubulovillous/villous adenomas (TV), and sessile serrated lesions (SSL). The analysis is adjusted for age, sex, lesion size (mm), and anatomic location (proximal, distal/rectal).\\
\\
Souvik Seal et al.](/content/explore/c108338f-6c9a-40ff-aa24-f6ef74d1b335?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1216519/tree/v1/index.html)

[Bioinformatics\\
\\
8 \| May \| 2026\\
\\
BRAPH 2: Reproducing Core-Periphery Brain Network Comparison Panels\\
\\
This capsule provides a MATLAB-based reproducibility environment for reproducing the Figure 2 comparison panels from the BRAPH 2 manuscript. It uses BRAPH 2 pipeline files (.b2) to regenerate anatomical, functional, and anatomical–functional multiplex core-periphery comparison brain-surface visualizations, and exports the corresponding significant brain-region tables. The capsule runs in MATLAB R2024a in a headless Code Ocean environment and saves all generated figures and Excel outputs to the results folder.\\
\\
Io̍k-uí Tiunn](/content/explore/b154069e-c50e-4357-8d9b-d687dd4e439f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6589176/tree/v1/index.html)

[Computer Science\\
\\
11 \| May \| 2026\\
\\
Acquisition-Parameter Effects on Image Utility in Near-Infrared Dorsal Hand Imaging: Linking Image Signatures to Repeatability and Separability\\
\\
This compute capsule accompanies a study evaluating acquisition-parameter effects on image-derived signatures and protocol-level image utility in near-infrared dorsal hand imaging. The capsule provides analysis notebooks, reproducibility notes, synthetic/template input files, and manuscript-supporting numerical outputs. Raw participant images, standardized ROI images, and ROI-level biometric data are not included because dorsal hand images constitute biometric information and are restricted by ethics and privacy constraints.\\
\\
Emre CANAYAz](/content/explore/7bdb4641-b4ae-473c-8c11-7c1dd5295e86?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6387831/tree/v2/index.html)

[Physics\\
\\
8 \| May \| 2026\\
\\
Asymmetric atmosphere drives phase-dependent CO absorption in WASP-121 b\\
\\
Code for generating and fitting light curves from JWST/NIRSpec G395H time series observations with a rotational transit model. For a demonstration, this capsule fits the white light curves obtained from a FIREFLY reduction of the observations using a least-squares approach. The fit to the data is plotted in results/white/leastsq/simultaneous\_light\_curve\_fit.pdf. The run time should be <1 minute. To fit the full rotational light curve model, set both the ROTATION\_1 and ROTATION\_2 parameters in the parameters.py file to True. To set R\_2/R\_\*=0, set ROTATION\_2 to False. To fit a translational light curve model, set both ROTATION\_2 and ROTATION\_1 to False.\\
If you would like to run the full pipeline including deriving the phase-resolved transmission spectrum, set the files2run parameter in 00\_run-file.py to \[1,2,3,4,5,6,9,7,8,10,11,12,13,14,15,16\]. For running the code on a different data set, replace the files in the data folder with your own 1D spectra. No special hardware is needed to run this code.\\
\\
Cyril Gapp](/content/explore/d20b98e1-ec67-44b5-b07f-7246c8154aa2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8754052/tree/v1/index.html)

published in [Nature Astronomy](/content/explore?query=Nature%20Astronomy&refine=journal/index.html)

[Engineering\\
\\
8 \| May \| 2026\\
\\
Measuring multi-site pulse transit time with an AI-enabled mmWave radar\\
\\
Pulse Transit Time (PTT) is a measure of arterial stiffness and a physiological marker\\
associated with cardiovascular function, with an inverse relationship to diastolic blood pressure (DBP). We present the first AI-enabled mmWave system for contactless multi-site PTT measurement using a single radar. By leveraging radar beamforming and deep learning algorithms our system simultaneously measures PTT and estimates diastolic blood pressure at multiple sites. The system was evaluated across three physiological pathways - heart-to-radial artery, heart-to-carotid artery, and mastoid area-to-radial artery achieving correlation coefficients of 0.73-0.89 compared to contact-based reference sensors. Furthermore, the system demonstrated correlation coefficients of 0.90-0.92 for estimating DBP, and achieved a mean error of -1.00-0.62 mmHg and standard deviation of 4.97-5.70 mmHg, meeting the FDA's AAMI guidelines for non-invasive blood pressure monitors. These results suggest that our proposed system has the potential to provide a non-invasive measure of cardiovascular health across multiple regions of the body.\\
\\
Jiangyifei Zhu et al.](/content/explore/c0d94c89-1834-47d5-9360-02d892b7fce4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6725422/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Bioinformatics\\
\\
7 \| May \| 2026\\
\\
SNPic: SNP Topic Modeling for Interpretable Clustering of Complex Phenotypes\\
\\
We present SNPic, a probabilistic framework that redefines the analytical landscape of complex trait genetics. By conceptualizing genetic associations as a highly structured probabilistic language, SNPic deconstructs fragmented, biobank-scale GWAS catalogs into an interpretable lexicon of \`\`genetic topics''. These inferred topics serve as fundamental, reusable biological modules that successfully dismantle rigid clinical boundaries, exposing the true pleiotropic spectrum of human pathology. Validated through rigorous mathematical simulations, stability-optimized inference on massive human cohorts, and generalization across diverse plant and animal species, our findings demonstrate that a generative, mixed-membership approach is essential for capturing the interconnected reality of the genome. Ultimately, SNPic shifts the field from cataloging isolated variants toward reconstructing an interpretable knowledge graph of the genome's latent semantic architecture. By providing a highly scalable, privacy-preserving, and biologically transparent analytical lens, SNPic establishes a powerful new cornerstone for integrative genomics, paving the way for next-generation patient stratification and precision medicine. This work also builds a conceptual bridge between two traditionally separate fields: statistical genetics and NLP, suggesting that advances in one domain may directly transfer methodological innovations to the other.\\
\\
Zhang Leyi et al.](/content/explore/92c0d7be-9844-4488-9277-ac7f4f7e620c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8324245/tree/v1/index.html)

[Engineering\\
\\
7 \| May \| 2026\\
\\
VibraHybrid-FD: An Open-Source Python Toolkit for Hybrid Feature Extraction and Selection in Vibration-Based Fault Diagnosis\\
\\
VibraHybrid-FD is an open-source Python toolkit for vibration-based fault diagnosis in rotating machinery. It combines hybrid time-frequency analysis — Short-Time Fourier Transform (STFT) for linear spectral features and Hilbert-Huang Transform/Empirical Mode Decomposition (HHT/EMD) for nonlinear adaptive decomposition — to extract a deterministic 56-dimensional feature vector from triaxial accelerometer signals.\\
The toolkit includes an optional Boruta-based feature selection module that reduces the feature set to approximately 15–20 statistically significant predictors, improving interpretability and reducing overfitting risk. A built-in benchmarking framework evaluates six classifiers (MLP, Logistic Regression, Quadratic SVM, Random Forest, CatBoost, LightGBM) and generates accuracy, F1-score, confusion matrices, ROC curves, and SHAP-based explainability plots.\\
\\
Dataset: Cooling Fan Motor Data (triaxial accelerometer, 9 combined fault/speed labels)\\
\\
Dependencies: Python 3.8+, numpy, pandas, scipy, scikit-learn, boruta, catboost, lightgbm, shap, emd, matplotlib, seaborn\\
\\
Tuan Minh Le et al.](/content/explore/acba5c7b-a1d9-4fe3-9646-58529aecf23d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2547605/tree/v1/index.html)

[Computer Science\\
\\
6 \| May \| 2026\\
\\
An open-source tool for simulating cyberattacks on smart grid datasets\\
\\
The source code aims to simulate both poisoned and unpoisoned attacks. The poisoned source code takes labeled or unlabeled training data as input and applies poisoning to generate adversarial samples in the training set using Projected Gradient Descent (PGD). This modified training set is then used for adversarial training.\\
\\
The unpoisoned source code aims to inject stealthy attacks, namely False Data Injection (FDI) attacks if the attack vector is available; otherwise, random false data injection is performed. It also injects three replay attacks. The generated dataset is called unpoisoned because each sample is correctly labeled as either attack or benign.\\
\\
KADRI Belkacem](/content/explore/f7d41f6e-82df-41a4-beb9-841a80542794?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4320426/tree/v1/index.html)

[Engineering\\
\\
6 \| May \| 2026\\
\\
Model Predictive Control of a Hybrid Thermal Management System\\
\\
MPC has gained popularity for its ability to satisfy constraints and guarantee robustness for certain classes of systems. However, for systems whose dynamics are characterized by a high state dimension, substantial nonlinearities, and stiffness, suitable methods for online nonlinear MPC are lacking. One example of such a system is a vehicle thermal management system (TMS) with integrated thermal energy storage (TES), also referred to as a hybrid TMS. Here, hybrid refers to the ability to achieve cooling through a conventional heat exchanger or via melting of a phase change material (PCM), or both. Given increased electrification in vehicle platforms, more stringent performance specifications are being placed on TMS, in turn requiring more advanced control methods. In this article, we present the design and real-time implementation of a nonlinear model predictive controller with 77 states on an experimental hybrid TMS testbed. We show how, in spite of high dimensions and stiff dynamics, an explicit integration method can be obtained by finding a suitable linear system at each time step within the MPC horizon online. This integration method further allows the first-order gradients to be calculated with minimal additional computational cost. Through simulated and experimental results, we demonstrate the utility of the proposed solution method and the benefits of TES for mitigating highly transient heat loads.\\
\\
Demetrius Gulewicz, Uduak Inyang-Udoh, Trevor Bird & Neera Jain](/content/explore/c4978618-2f46-4987-9693-f3208ef3631d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8013612/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tcst.2025.3646704) published in [IEEE Transactions on Control Systems Technology](/content/explore?query=IEEE%20Transactions%20on%20Control%20Systems%20Technology&refine=journal/index.html)

[Physics\\
\\
6 \| May \| 2026\\
\\
Meta-optimization of maximally localized Wannier functions\\
\\
**General Information** \\
\\
This capsule runs automatic Wannierization. Below we describe the code and the functioning\\
\\
**Code** \\
\\
The code is stored in the "/code" directory of the capsule. It is contained inside the folder "AutoWann".\\
Here, autowann.py is the main module which calls other modules such as autowann\_optimizer.py (which has the objects and methods for optimization), and util module which has the utilities needed for performing loss function minimization. autowann.py calls scikit-optimize, bayesian-optimization, and scipy optimization algorithms in the library mode. Further, it calls EPWpy to build the workflow for Wannierization on top of which optimization is performed. \\
Here is a brief description of each module\\
\\
autowann.py: Main module for automatic Wannierization. Here, we define all parameters needed for wannierization alongside defining the optimization class optained from autowann\_optimizer.py.\\
\\
autowann\_optimizer.py: Contains the functions that perform the optimization of the loss functions for Wannierization. loss functions are defined in 'get\_opjective". Here loss function "median" defines the L1 loss function from text. 'median\_no\_hev" defines L2 loss function from the main text, "elph\_decay" defines the L3 loss function from the main text, and "median\_balanced" defines the L4 loss function.\\
\\
util.band\_distance.py: returns the bandstructure related attributes.\\
util.elph\_util.py: returns electron-phonon matrix element related attributes.\\
util.wannier\_util.py: returns the wannier function related attributes.\\
util.permute\_opb.py: generates the orbitals for Wannierization.\\
util.default.py: default parameters for Wannierization.\\
\\
**Example Run**\\
\\
- In this example, we obtain the Wannier orbitals for single-Layer MoS2 using our method. Here, for the simplicity and resource\\
\\
restriction (only 4 cores), we do not account for spin-orbit-coupling(SOC). Also, the loss function is a simple loss function which minimizes the bandstructure interpolation error with a constraint on the spread (10 angstrom^2). Here the interpolation error ((interp\_max+interp\_avg)/2) is an average between the maximum interpolation error (interp\_max) and the average interpolation error (interp\_avg). The criterion for the loss function is set to 0.025 eV which is controlled using "hard\_tol" parameter. This is a loose criterion and should be much smaller for a production calculation.\\
\\
- In this example, we perform a full cylce of a typical calculation (1) self-consistent-field (SCF), (2) Non-SCF (NSCF), (3) Phonon (because Wannierization is performed inside EPW and EPW needs a phonon calculation to start Wannierization), (4) Bandstructure calculation on a path using DFT, (5) Wannierization using Wannier90 in library mode using EPW.\\
\\
- The details for all these four steps (and computational parameters) are provided in "example.py" where various parameters can be tuned. example.py also contains all information regarding parameters for automated Wannierization which the Referee can tune.\\
\\
- Finally, once the optimization is reached where the Loss function (Obj\_func) value is below the "hard\_tol", the calculation stops returning the Bestcase.npy.\\
\\
- The outputted Bestcase.npy is rerun using example\_5.py to obtain the bandstructure for comparison. \\
\\
- Finally the bandstructure is plotted using band\_plot.py. \\
\\
- All these steps are performed automatically by clicking "Reproducible Run" button on the top right. Once the calculation finishes, the results are stored in "/results" folder.\\
\\
\\
**Results** \\
\\
The results of the calculation are provided in the "/results/" folder.\\
In the "/results/" folder, the bandstructure from optimization is provided in "Bandstructure.png" where the bandstructure obtained from AutoWann is provided as dotted blue line while the bandstructure from DFT is provided as solid black lines. \\
The meta-optimization steps are provided in the "Metadata.npy".\\
While the best optimization result is stored in "Bestcase.npy".\\
The raw output file is provided in the "output" file, where the information regarding the optimization can be infered.\\
Eventually, the user must obtain a bandstructure similar to the one shown in "data/Bandstructure.png". It should be noted that the match is only expected within the shaded yellow region which is used for optimization and controlled using "E\_min" and "E\_max" parameters in the "example.py".\\
\\*\\* An example of optimized output\\
{'string': \["'Mo: l= -3, mr= 1,4'", "'S: l= 3, mr= 1,2,3,5'"\], 'num\_wan': 10, 'diswin': \[-8.56166111771633, -1.5726501320620683, -16.749067524559987, 7.272919354255732\], 'spread\_sum': 21.28282147, 'num\_bands': 24, 'median\_spread': 1.66969637, 'Obj\_func': 0.01612626079189362, 'iterations': 7, 'Obj\_L2': 0.03178309999999973, 'time': 76.50406098365784}\\
\\
Here, \\
"string" is the Wannier projections obtained from AutoWann, \\
"num\_wan" is the number of Wannier function, \\
"diswin" shows the disentanglement windows \[dis\_froz\_min,dis\_froz\_max,dis\_win\_min,dis\_win\_max\], \\
"spread\_sum" is to total spread in angstrom^2, \\
"median\_spread" is the median spread in angstrom^2, \\
"num\_bands" is the number of DFT bands used for Wannierization,\\
\\
"Obj\_func" is the objective function value, \\
"Obj\_L2" is the L2 norm of band interpolation error,\\
"iterations" is total DE iterations,\\
"time" is the time taken in seconds for finding this solution.\\
\\
**Downloading results** \\
\\
The Referee can download the result and check the above parameters in their computer system. This can be performed by downloading the "Bestcase.npy" generated in the "/results" and then using a simple python code written below (open\_best.py).\\
\\
open\_best.py\\
#--------------------------------------------------#\\
import numpy as np\\
A=np.load('Bestcase.npy',allow\_pickle=True).item()\\
print(A)\\
#--------------------------------------------------#\\
\\
Sabyasachi Tiwari, Bruno Cucco, Viet-Anh Ha & Feliciano Giustino](/content/explore/28f5ed1a-ac07-4007-91d1-2c3c596d5dbc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3948614/tree/v1/index.html)

[Associated article](https://doi.org/10.1038/s41524-026-02082-1) published in [npj Computational Materials](/content/explore?query=npj%20Computational%20Materials&refine=journal/index.html)

[Engineering\\
\\
6 \| May \| 2026\\
\\
Unified Single Pass Multi-Output Streaming Data-flow Architecture\\
\\
This capsule enables verification of the functional aspects of the proposed research work. The capsule includes a software-based reference implementation, representative input image, and scripts to reproduce the key results and figures reported in the manuscript.\\
\\
Fakhra Aftab](/content/explore/147b619a-7c91-45a0-9664-28b9a9ec30fd?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1780698/tree/v1/index.html)

[Computer Science\\
\\
15 \| May \| 2026\\
\\
Interpretable Visual Recognition using Semi-supervised Sparse Autoencoders\\
\\
A Semi-Supervised Explainable Visual Recognition Model Based on SAE\\
\\
Yinsheng Zhang & Xudong Yang](/content/explore/81e62071-01b1-455a-8364-25b9a4279658?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5905638/tree/v1/index.html)

[Computer Science\\
\\
6 \| May \| 2026\\
\\
Low field MRI data reconstruction demo\\
\\
Magnetic resonance imaging (MRI) is a versatile tool for medical diagnosis of soft tissue. Its drawbacks include the long scan time, expensive hardware and non-portability. The use of low magnetic field enables the development of smaller, portable, more affordable hardware. This has the potential to enable widespread adoption of MRI for rapid diagnosis and delivery of treatment in emergencies. Unfortunately, the hardware advantages come at the cost of reduced resolution, low signal to noise ratio, and vulnerability to outside electromagnetic interference. This leads to longer scan times and lower image quality. In this work we propose a deep learning based method for the enhancement of low-field MRI scans while allowing scan acceleration. Our method is able to work under various noise levels, and produces a distribution of reconstructions to allow visualisation of uncertainty. Additionally, we compare the use of methods for adaptively optimising the scan sequence. We conduct experiments on simulated data from the fastMRI dataset and show the results of the proposed approach. We are able to generate a distribution of high resolution 256×256 reconstructions from noisy 64× 64 scans at 8dB to 15dB SNR, improving the image PSNR.\\
\\
Amr Morssy & Paul D. Teal](/content/explore/931ccb0b-a20c-4599-aa38-a3f4df9b9fd3?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9933713/tree/v1/index.html)

[Medical Sciences\\
\\
5 \| May \| 2026\\
\\
Subsample-Shift Estimation Based on Even Convex Combinations in Dual-Layer Flat-Panel Detector Alignment\\
\\
Conduct subsample-accuracy shift estimation to align x-ray images acquired from dual-layer flat-panel detectors (DFDs). Example slat-edge and chest x-ray images acquired from a DFD prototype are included.\\
\\
Dong Sik Kim](/content/explore/56b9df4a-6f30-46ef-9254-843e0021da85?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0592250/tree/v1/index.html)

[Computer Science\\
\\
5 \| May \| 2026\\
\\
Aurea Sim\\
\\
Reproducible compute capsule containing the experimental data and semantic variance analysis for AureaSim.\\
\\
Robert Waszkowski](/content/explore/42526fc6-4592-492c-aebb-d0b39a037287?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7502577/tree/v2/index.html)

[Bioinformatics\\
\\
22 \| May \| 2026\\
\\
AI4EVER v2.3: Tabular Genomic Prediction Workflows](/content/explore/eb06fcbb-ff87-45ea-8f42-582cd965bec3?page=2&filter=all/index.html)

[This capsule provides a reproducible demonstration of the AI4EVER v2.3 tabular genomic prediction backend. AI4EVER is an interactive graphical platform for machine learning and deep learning in agricultural and biological data analysis.\\
The capsule includes example genotype, phenotype, and GWAS datasets and demonstrates single-trait and multi-trait genomic prediction workflows with and without GWAS-derived marker weighting. Five models are included: Ridge regression, Random Forest, Gradient Boosted Decision Trees, scikit-learn MLP, and TensorFlow/Keras neural network.\\
The default run executes the complete reviewer workflow: environment check, multi-trait training without GWAS, multi-trait training with GWAS, multi-trait saved-model prediction, single-trait training without GWAS, single-trait training with GWAS, and single-trait saved-model prediction. Outputs include model files, summary tables, prediction tables, cross-validation scatterplots, and contact-sheet comparison figures saved under `/results`.\\
The prediction-only scripts reuse the example genotype file because no independent external genotype dataset is included. This step demonstrates the saved-model prediction workflow and should not be interpreted as independent external validation.\\
The full AI4EVER source code is available at:](/content/explore/eb06fcbb-ff87-45ea-8f42-582cd965bec3?page=2&filter=all/index.html) [https://github.com/liangmeijing89/AI4EVER](https://github.com/liangmeijing89/AI4EVER)

Meijing Liang, Yang Hu & Zhiwu Zhang

[Open Capsule](/content/capsule/3041436/tree/v2/index.html)

[Mathematics\\
\\
4 \| May \| 2026\\
\\
RH via E₈ Dirac Index — Formal Verification\\
\\
Title: "RH via E₈ Dirac Index — Formal Verification"\\
DOI: 10.5281/zenodo.20005292\\
GitHub: github.com/Heime-Jorgen/rh-e8-lean\\
\\
Peter Jang](/content/explore/c106ccc1-f622-445f-8ae1-302f70885834?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0949201/tree/v1/index.html)

[Engineering\\
\\
4 \| May \| 2026\\
\\
An Elite Circuit Model-Enhanced Frequency Control for Self-Resonant PCB-Based Hybrid Wireless Power Transfer Systems\\
\\
The code is developed for parameter identification of the HPT coupler.\\
\\
Junxiang Yang, Kaiyuan Wang & Sun Zhen](/content/explore/ab0f1838-8fe8-42eb-862b-8053d0398741?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5142876/tree/v1/index.html)

[Mathematics\\
\\
4 \| May \| 2026\\
\\
Quantifying uncertainty in drift diffusion models of decision making\\
\\
Decision-making behavior changes over time, exhibiting temporal correlation and nonstationarity. Existing drift diffusion model (DDM) fitting methods either do not provide uncertainty of parameter estimates, or rely on restrictive assumptions that decisions are independent and that parameters remain constant over time, potentially underestimating the uncertainty. To address these limitations, we propose a computationally efficient method for estimating analytical uncertainties in DDM parameters that are robust to temporal dependence and unmodeled parameter variability, while explicitly modeling nonstationary variability through covariates. We apply this method to rat decision-making in a two-alternative forced-choice visual task, revealing dynamic decision-making states across multiple timescales.\\
\\
Gabriel Riegner, Armin Schwartzman & Pamela Reinagel](/content/explore/1ef431d1-462f-493f-b370-5296c25755f3?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4722579/tree/v1/index.html)

[Physics\\
\\
4 \| May \| 2026\\
\\
Time-Field Spectral Curvature and Stability-Selected Closure: A Minimal Reproducible Model for Emergent Gravitational Coupling\\
\\
This capsule provides a minimal, reproducible numerical implementation of the framework introduced in:\\
\\
“Newton’s Constant as a Time-Field Spectral Curvature: A Wheeler–DeWitt Closure Pathway”\\
\\
The model demonstrates how gravitational coupling can be recast as an emergent quantity arising from the spectral structure of a scalar time-field. Rather than treating Newton’s constant G as a fundamental input, the framework reduces it to a function of:\\
\\
the curvature of an effective time-field potential 𝑉eff′′​(Θ0​),\\
the intrinsic correlation length of the time-field Hamiltonian,\\
and a stability-selected closure condition defined by a universal dimensionless constant χ.\\
\\
Matthew J. Hall (0009-0001-7066-2558)](/content/explore/49c1ace7-c579-4728-88a8-6def539caf25?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7324725/tree/v1/index.html)

[Computer Science\\
\\
4 \| May \| 2026\\
\\
Supplement for: Mining Agent Reasoning Strategies: Process Discovery and Signal Decomposition for LLM Trajectory Behavior Analysis\\
\\
This capsule cold-start reproduces the data products reported in the\\
anonymous IEEE TAI submission "Mining Agent Reasoning Strategies: Process\\
Discovery and Signal Decomposition for LLM Trajectory Behavior Analysis".\\
\\
The capsule is intentionally separated from the LaTeX manuscript source. It\\
extracts source trace archives, rebuilds event logs, recomputes the signal\\
decomposition and behavior metrics, and regenerates the computational tables,\\
CSV summaries, and figure assets used to support the paper's empirical\\
claims. The run is fully headless and writes all generated artifacts to\\
`/results`.\\
\\
Main outputs include:\\
\\
- `figures/*.pdf` and `figures/*.png` for the data figures in the paper\\
- `tables/*.tex` and `tables/*.csv` for the reported tables\\
- `summary_metrics.json` and `tai_reproducibility_manifest.json`\\
\\
The `/data/source_archives` directory contains the source trace archives and\\
`/data/models` contains the local embedding models used offline. The only\\
manual exception is `/data/manual_audit`, which contains human audit labels\\
and notes; all scores, counts, figures, and tables are recomputed.\\
\\
Anonymous](/content/explore/836577df-cb4b-40a2-9adc-e06abcbf02f1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6750940/tree/v1/index.html)

[Computer Science\\
\\
4 \| May \| 2026\\
\\
Gap-Aware Candidate Reranking for Diffusion-Guided Multi-Strategy Floorplanning\\
\\
This is the codebase to reproduce "Gap-Aware Candidate Reranking for Diffusion-Guided Multi-Strategy Floorplanning"algo for better Physical Design of Chips\\
\\
SHASHANK](/content/explore/b9343345-070a-472a-86d8-056ea997aa1d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0072243/tree/v1/index.html)

[Bioinformatics\\
\\
2 \| May \| 2026\\
\\
Spatial transcriptomics of TNBC across clinical states reveals subtype-specific wiring and immunosuppressive niches\\
\\
The capsule includes all data and code required to reproduce the analysis. The data files are located in the “data” folder, and the scripts are in the “code” folder. To reproduce the analysis, right-click the R script and select “Set as File to Run.”\\
\\
Fengyuan Huang, Nitish Kunte, Himani Khurana & Clayton Yates](/content/explore/afcd3981-d0ac-406b-8c72-6a499df8f197?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4733243/tree/v1/index.html)

[Computer Science\\
\\
1 \| May \| 2026\\
\\
LUXNet: Light Field View Synthesis using Cross-Domain Feature Interaction and Spatial-Angular-Depth Correlation\\
\\
Dense Light Field (LF) view synthesis is highly desirable for applications such as 3D scene reconstruction, depth \\
estimation, and post-capture refocusing. However, existing hardware-based dense LF acquisition methods are either costly or \\
limited by sensor resolution, which imposes a trade-off between angular and spatial resolution. To overcome these limitations, \\
various learning-based depth-dependent Light Field View Synthesis (LFVS) methods with large receptive fields have been \\
proposed. However, these methods often fail to model fine details and to handle complex occlusions effectively. In contrast, nondepth-dependent LFVS methods typically overlook geometric priors, resulting in the loss of fine textures and angular consistency.\\
To address these issues, we propose LUXNet, a novel LFVS method that uses cross-domain feature interaction and spatial-angulardepth correlation to synthesize dense LFs. The LUXNet method consists of four modules: (i) Shallow Feature Extraction (SFE), \\
(ii) Deep Feature Extraction (DFE), (iii) Spatial-Angular-Depth Correlation (SADC), and (iv) Angular Upsampling (AU). The \\
SFE module projects sparse input views into a high-dimensional space for discriminative feature learning. The DFE module uses \\
three stacked Multi-Scale Residual Blocks (MSRBs) to extract dense features from multiple representations. The extracted features \\
are then fused into a unified cross-domain feature representation and passed to the SADC module to enforce angular consistency. \\
Finally, the AU module upsamples the micro-image feature representation to synthesize dense LFs. Experimental results on realworld and synthetic datasets demonstrate that LUXNet outperforms state-of-the-art methods in synthesizing high-fidelity dense \\
LFs.\\
\\
Muhammad Zubair, Caroline Conti,, Paulo Nunes & Luís Ducla Soares](/content/explore/78ae497e-aec4-4672-b9b9-a867b8f9608f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6367121/tree/v1/index.html)

[Bioinformatics\\
\\
1 \| May \| 2026\\
\\
Large Language Model Agents for Evidence Based Genetic Disease Severity Classification\\
\\
**Background** \\
\\
Disease severity classification for genetic conditions is subjective and labor intensive, creating bottlenecks in genomic screening where commercial panels vary widely in size and overlap.\\
\\
**What this work does** \\
\\
We developed an autonomous AI agent that integrates **Reasoning and Acting (ReAct)** with **Retrieval-Augmented Generation (RAG)** to classify **10,211 Human Phenotype Ontology (HPO) terms**. The agent applies American College of Medical Genetics ( **ACMG**) endorsed severity guidelines and American College of Obstetricians and Gynecologists ( **ACOG**) quality of life criteria, retrieves supporting evidence from PubMed, generates interpretable reasoning chains, and independently verifies every cited claim against its source abstract. Phenotype level tiers are then aggregated into gene level severity profiles and compared against established reproductive carrier and secondary findings panels. The system is designed to support standardized panel design by providing reliable, automated severity classification grounded in retrievable, verifiable evidence.\\
\\
**What this capsule reproduces** \\
\\
Running this capsule reproduces, end to end:\\
\\
- **Per term classifications, reasoning traces, and source verification scores** (`main.py`): the full ReAct + RAG agent that produced the manuscript's classifications.\\
- **Figure S1** (`hyperparameter.py`): model x temperature accuracy and response time heatmap.\\
- **Gene level severity analysis** (`severity_classification.py`): the Profound / Severe / Moderate / Mild distribution per gene disease pair after the 30 percent frequency filter, body system breakdown, mode of inheritance breakdown, and external comparisons against the Mackenzie's Mission carrier panel and the ACMG Secondary Findings and Carrier Screening lists.\\
\\
**Reviewer reference code** \\
\\
`visualization.py` (Figures 2 to 6, S2, S3 and the validation dashboard) is included for reviewer code inspection but is not executed by `run`, because it depends on the IP protected classifier output `Classified_terms.csv`, which is available from the corresponding author under a Material Transfer Agreement with UNSW Sydney.\\
\\
Tohid Ghasemnejad et al.](/content/explore/eea3f49a-e54f-45c0-af11-1a60c5bcf39e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4872093/tree/v1/index.html)

[Computer Science\\
\\
30 \| Apr \| 2026\\
\\
Surrogate Explainability Pipeline for Audio Deepfake Detection\\
\\
This capsule provides a reproducible implementation of a modular pipeline for surrogate-based explainability of multimodal large language models for audio deepfake detection. It includes acoustic feature extraction with openSMILE, training a Random Forest surrogate model on these features with model predictions as targets, and global and local explainability analyses using SHAP and LIME. The system is configured via a single `config.yaml` file and executed through `run.py`, enabling end-to-end reproducibility of feature extraction, surrogate modeling, and interpretability analysis. Precomputed predictions from fine-tuned Gemma and Qwen models are included to ensure reproducibility without requiring GPU-intensive supervised fine-tuning, which was performed externally and is not required to run this capsule. The design reflects the modular architecture proposed in the paper and supports easy integration of additional models for comparative explainability analysis.\\
\\
Vishnu S. Pendyala & Janani Kripa Manoharan](/content/explore/1147f2df-0c12-4c6c-ac25-63c0844fa817?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5014210/tree/v1/index.html)

[Computer Science\\
\\
30 \| Apr \| 2026\\
\\
Supplementary code for Confidence-Guided Omnidirectional Stereo with Geometry-Regularized Refinement for Real-Time Fisheye Depth Estimation\\
\\
This capsule provides supplementary code for “Confidence-Guided Omnidirectional Stereo with Geometry-Regularized Refinement for Real-Time Fisheye Depth Estimation.” It includes source code, a pretrained checkpoint, and a small demo test set to verify the inference and evaluation pipeline. The reproducible run loads the checkpoint, performs depth prediction, and reports metrics. Full benchmark datasets are not included due to size and licensing restrictions, so the demo results are intended to validate the executable pipeline rather than reproduce all paper results.\\
\\
Yushen Wang](/content/explore/9adaa6d9-8b2e-46fa-be47-b772ed168476?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7748367/tree/v1/index.html)

[Computer Science\\
\\
29 \| Apr \| 2026\\
\\
CashewNet: An Ensemble Deep Learning Framework with Attention and Multi-Scale Feature Fusion for Cashew Leaf Disease Detection\\
\\
This work presents CashewNet, a novel ensemble deep learning framework specifically designed\\
to improve the accuracy and robustness of cashew leaf disease detection. The proposed model\\
integrates attention mechanisms and multi-scale feature fusion to effectively capture both finegrained and large-scale disease patterns. By combining multiple architectures and leveraging\\
advanced feature representation strategies, the framework achieves superior performance\\
compared to state-of-the-art models.\\
\\
Daudi Flavian & Sakthivel, R](/content/explore/ab424e23-29c8-4483-abf4-6e3ebf62425c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4788878/tree/v1/index.html)

[Bioinformatics\\
\\
29 \| Apr \| 2026\\
\\
Depletion of effector regulatory T cells associates with major response to induction dual immune checkpoint blockade\\
\\
Codes used to reproduce main figures in the manuscript.\\
\\
Xianli Jiang et al.](/content/explore/5342814d-14cb-4ecf-a807-7ae50013ecee?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7053274/tree/v1/index.html)

[Engineering\\
\\
28 \| Apr \| 2026\\
\\
A Probabilistic Generative Model for Spectral Speech Enhancement Codebase\\
\\
This repository accompanies the paper:\\
\\
M. Hidalgo-Araya et al., "A Probabilistic Generative Model for Spectral Speech Enhancement", 2025.\\
A comprehensive evaluation framework for virtual hearing aids using the VOICEBANK\_DEMAND dataset with warped filter bank (WFB) preprocessing.\\
\\
Marco Hidalgo-Araya et al.](/content/explore/63141d79-cad0-40f9-a612-8d0e897edd00?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2687089/tree/v1/index.html)

[Engineering\\
\\
6 \| May \| 2026\\
\\
Dynamic Time Warping for Secondary Path Interpolation in Local Active Noise Control\\
\\
For stable and performant local active noise control (ANC), accurate estimates of the paths between secondary sources and the desired points of cancellation are required. When these paths are recorded prior to operation for a limited number of positions, this poses a problem in scenes with moving points of cancellation. In this article, we propose an interpolation approach for filter coefficients of secondary paths. By applying dynamic time warping (DTW) in an offline analysis, impulse responses are properly aligned before interpolation and de-warping during operation. The system is benchmarked against nearest-neighbor and linear interpolation with and without global time alignment for translation and rotation of a listener's head. Our DTW-based technique exhibits lower overall system mismatch and extends the frequency range for stable operation, especially for lateral translation and coarse measurement grids. In simulations, higher noise reduction and extended frequency range of stable operation could be observed compared to the other baseline systems. The proposed technique can reduce the required number of measurement positions of secondary paths substantially while increasing the controlled frequency bandwidth for time‑variant real‑world systems.\\
\\
Felix Holzmüller & Alois Sontacchi](/content/explore/068921bb-ad16-488b-99ec-ad7f136ee6ea?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5970407/tree/v1/index.html)

published in [IEEE Open Journal of Signal Processing](/content/explore?query=IEEE%20Open%20Journal%20of%20Signal%20Processing&refine=journal/index.html)

[Computer Science\\
\\
28 \| Apr \| 2026\\
\\
CLAPC: A Hybrid CNN-LSTM Architecture for Automated Passenger Counting from Video Streams in Public TransportApr 25, 2026 06:49\\
\\
CLAPC — CNN-LSTM Automated Passenger Counting from 2D Video Streams\\
CLAPC is an automated passenger counting framework designed to estimate boarding and alighting passenger numbers from 2D video streams. The method combines a Convolutional Neural Network with a recurrent Long Short-Term Memory network. The CNN extracts spatial features from video frames, while the LSTM captures temporal information across frame sequences.\\
\\
The capsule supports both training and inference. \\
\\
Due to commercial data restrictions, the original 2D video dataset cannot be publicly released with this capsule. Instead, the capsule provides the publicly available Berlin-APC 3D video dataset for training. This dataset was also used in our paper and serves as an alternative dataset for demonstrating and reproducing the training workflow. For inference, the capsule includes sample 2D video data and two pretrained models. These models were trained using differently processed versions of the data, allowing users to run and compare inference results on the provided video samples.\\
\\
Sambu Seo](/content/explore/b5f84c44-ccba-4cd2-989f-9e7954f71311?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3565461/tree/v1/index.html)

[Computer Science\\
\\
8 \| Jun \| 2026\\
\\
Artifact: Spatiotemporal Robustness of Temporal Logic Tasks using Multi-Objective Reasoning\\
\\
This artifact accompanies the CAV 2026 paper: **"Spatiotemporal Robustness of Temporal Logic Tasks using Multi-Objective Reasoning" by Oliver Schön and Lars Lindemann (ETH Zürich).**\\
\\
It provides MATLAB implementations of the spatiotemporal robustness (STR) monitoring algorithms from the paper and reproduces the two case study figures (F-16 fighter jet and Waymo robotaxi scenario).\\
\\
Oliver Schön & Lars Lindemann](/content/explore/0fef0b07-cfbd-4c98-ada2-bf3ccb22437e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0448850/tree/v2/index.html)

[Computer Science\\
\\
27 \| Apr \| 2026\\
\\
Supplement for: Inductive Miner for ToolLLaMA: Semantic Process Comparison of Chain-of-Thought and Depth-First Search Traces\\
\\
**Capsule** \\
\\
This capsule reproduces the computational results reported in the paper:\\
\\
Inductive Miner for ToolLLaMA: Semantic Process Comparison of Chain-of-Thought and Depth-First Search Traces\\
\\
Using archived ToolBench trace data provided in `/data`, the capsule rebuilds the event logs, process-mining summaries, semantic comparison outputs, and paper assets used in the manuscript. The run is fully headless and writes all generated artifacts to `/results`.\\
\\
Main outputs include:\\
\\
- `evaluation_rows.csv` and `experiment_summary.csv` for split-level quantitative comparisons\\
- `generalization_direction_summary.csv` for cross-model robustness analysis\\
- `manual_arg_flow_audit_summary.json` and `matched_case_summary.csv` for validation analyses\\
- paper figure assets (`*.png`) and the main table source (`table1_main_comparison.tex`) under `paper_assets/`\\
\\
Ying Xu](/content/explore/ce3ee1d0-df58-4728-9034-98ba3a8cbae3?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2679705/tree/v1/index.html)

[Bioinformatics\\
\\
4 \| May \| 2026\\
\\
Leveraging Computational Methods for Bioinformatics and Drug Discovery\\
\\
This capsule contains the computational pipeline code associated with the paper 'Leveraging Computational Methods for Bioinformatics and Drug Discovery'. The code implements ESM-1b protein embedding simulation, PCA and UMAP dimensionality reduction, k-means clustering, and evaluation metrics including Silhouette Score, Davies-Bouldin Index, and Adjusted Rand Index.\\
\\
Amer Majzoub, Anselmo Tomás Fernández Pena](/content/explore/caaa9b4b-3de2-4fd5-ae9d-b39a6bb83253?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1015191/tree/v1/index.html)

[Engineering\\
\\
27 \| Apr \| 2026\\
\\
Compartmental modeling for the separation of dual-tracer PET within a preclinical model triple negative breast cancer\\
\\
This code helps in fitting compartmental modeling for dynamic dual tracer PET. This script performs voxel-wise fitting of dual-tracer PET data using compartmental models.\\
\\
- It processes dynamic PET images of two tracers injected sequentially, estimates kinetic parameters voxel-wise, and generates parametric maps. The script supports both individual and population-based\\
- arterial input functions (AIFs), fits the AIFs to exponential models, and uses the fitted AIFs for kinetic modeling. The main steps include:\\
  - Reading in dynamic PET and CT images and masks.\\
  - Selecting the AIF method (individual or population).\\
  - Fitting the AIF curves to exponential models.\\
  - Performing voxel-wise fitting using compartmental models in parallel.\\
  - Calculating SUVs, kinetic parameters, and R² maps.\\
  - Saving and exporting the results.\\
  - Computing and saving statistical summaries for tumor variables.\\
\\
Urvi Rawal, Ameer Mansur, Carlos Gallegos & Lily Watts](/content/explore/1c5deebc-add6-40c6-9aa7-5c1f9d1a650d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6606776/tree/v1/index.html)

[Engineering\\
\\
24 \| Apr \| 2026\\
\\
Neural Rate-Adaptive LDPC Decoding for the Slepian-Wolf Problem\\
\\
Distributed source coding (DSC) enables efficient compression of correlated sources by performing independent encoding and joint decoding. Slepian-Wolf coding plays a central role in DSC, as it allows near-lossless compression of correlated data at rates asymptotically approaching the joint entropy. Traditional approaches to constructive Slepian-Wolf coding employ algebraic binning techniques using channel codes, with decoding performed via algorithms such as belief propagation (BP). While effective, these methods suffer from high computational cost and suboptimal compression performance. We propose the first constructive learned Slepian-Wolf decoder for rate-adaptive coding, using a single multi-rate Transformer model. The architecture is inspired by neural channel decoding, but addresses the unique challenges of syndrome-based Slepian-Wolf coding based on side information. Furthermore, we incorporate the proposed rate-adaptive neural Slepian-Wolf decoder into a novel neural layered Wyner-Ziv code design for the quadratic Gaussian case and into a new layered Wyner-Ziv design for distributed stereo image coding. For Slepian-Wolf coding of binary sources, our neural decoder improves the compression performance over traditional BP decoding by up to 11%. In our monolithic and layered Wyner-Ziv designs, we are between 0.05 and 0.2 bits/sample away from the estimated ideal rate bound, while entropy coding needs an additional 0.4-1.2 bits/sample compared to the proposed Slepian-Wolf codec. Moreover, our stereo image coding design reduces the coding rate by 9-19% in low rate settings compared to the state-of-the-art with minimal loss in image quality. Finally, the proposed decoder is 15~times faster than BP decoding on a GPU.\\
\\
Brent De Weerdt & Nikos Deligiannis](/content/explore/80dd6503-0ce8-488a-b208-63e692b4b027?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7034976/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/ojsp.2026.3684426) published in [IEEE Open Journal of Signal Processing](/content/explore?query=IEEE%20Open%20Journal%20of%20Signal%20Processing&refine=journal/index.html)

[Bioinformatics\\
\\
24 \| Apr \| 2026\\
\\
Population genomics of Cryptosporidium bovis in China reveals drivers of evolution and adaptation\\
\\
Cryptosporidium bovis is a common parasite infecting post-weaned calves. It differs from the zoonotic species Cryptosporidium parvum in terms of host range, age distribution, and pathogenicity. Therefore, genomic data from C. bovis are useful for understanding the evolution of host infectivity and virulence of Cryptosporidium spp. In this study, we sequenced the genome of C. bovis in 23 samples from China, covering subtypes XXVIa to XXVIf. We constructed a near-chromosome-level reference genome for C. bovis, enabling the characterization of structural variations between C. bovis and C. parvum. Phylogenetic analyses revealed that C. bovis isolates cluster by farm rather than by subtype. Within-farm recombination and mixed infections further support the limited utility of single-locus subtyping. Comparative genomic analyses identified 38 highly polymorphic genes, including 18 under positive selection. This study reveals that sympatric recombination drives genomic homogenization in C. bovis. These findings underscore its potential as an indicator species, as its population structure reflects local infection pressure and transmission risk. These insights collectively enhance our understanding of Cryptosporidium evolution and transmission dynamics.\\
\\
Tianyi Hou](/content/explore/bdce2c13-5952-4518-a342-50c35fd14633?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1875881/tree/v1/index.html)

[Computer Science\\
\\
24 \| Apr \| 2026\\
\\
Privacy-Preserving Protocols for Machine Learning: Design and Application to Neural Network Training (P3ML)\\
\\
Existing Secure Multi-Party Computation (SMPC) frameworks for privacy-preserving machine learning often suffer from prohibitive communication overhead and inefficiency in handling non-linear layers, particularly Batch Normalization (BN). To address these bottlenecks, this paper introduces P3ML, a high-efficiency privacy-preserving framework tailored for Neural Network (NN) training in a semi-honest three-party setting. P3ML systematically optimizes core building blocks by introducing a novel fixed-point multiplication truncation protocol that eliminates offline precomputation and reduces bandwidth. To accelerate non-linear activation, a constant-round protocol for the derivative of ReLU (DReLU) is designed, requiring only two communication rounds. Moreover, addressing the normalization gap, P3ML proposes a secure BN protocol leveraging bit decomposition and cumulative OR operations to achieve logarithmic round complexity. Formal security analysis demonstrates the framework's robustness against semi-honest adversaries under an honest-majority assumption. Extensive experiments on real-world datasets reveal that P3ML reduces total communication volume by approximately 50% and accelerates training speed by 3 to 5 times compared to state-of-the-art baselines such as SecureNN and FALCON, establishing its practical efficacy for complex deep learning architectures.\\
\\
Xing Zhang, Zezhou Fei, Chenyang Shao & Changda Wang](/content/explore/57b87b8e-34d7-4961-836f-3b75119a88ba?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7948154/tree/v1/index.html)

[Physics\\
\\
24 \| Apr \| 2026\\
\\
superconducticity from fluctuating loop current calculation\\
\\
gap equation, band structure, and cooper vertex plot\\
\\
Daniel Schultz et al.](/content/explore/6c5d94ec-04d9-4fc8-b8e5-c0a347035804?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6354792/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Social Sciences\\
\\
23 \| Apr \| 2026\\
\\
Replication Data for: Exceptional Measures for Exceptional Times: Deciphering Emergency Politics in the Council of the European Union\\
\\
This repository includes only the code to replicate the main analyses from the paper "Exceptional Measures for Exceptional Times: Deciphering Emergency Politics in the Council of the European Union."\\
\\
Paula Montano](/content/explore/27fc4ddd-7305-41b8-8e65-4d51ff51f3f5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2786807/tree/v1/index.html)

[Computer Science\\
\\
23 \| Apr \| 2026\\
\\
Why You Shouldn't Use Tdoa for Multilateration\\
\\
The maximum likelihood problem arising from multilateration or source localization via signal times of arrival (TOA) leads to a nonlinear least squares problem in target position and target transmission time (TTT). Since we are not interested in the latter, it is usually eliminated from the equation system. We eliminate the TTT in closed form, which is simpler to design, easier to implement, and faster to compute than the often used pairwise time differences of arrival (TDOAs). We propose an unweighted nonlinear least squares formulation of the multilateration problem whose minimization with the Levenberg-Marquardt algorithm is very fast.\\
\\
Daniel Frisch & Uwe D. Hanebeck](/content/explore/7eac463e-5da2-4f16-85a0-36000b03c5f5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5865722/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/mfi67357.2025.11259346) published in [2025 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (MFI)](/content/explore?query=2025%20IEEE%20International%20Conference%20on%20Multisensor%20Fusion%20and%20Integration%20for%20Intelligent%20Systems%20(MFI/index.html)&refine=journal)

[Bioinformatics\\
\\
23 \| Apr \| 2026\\
\\
AGCECDA: Attention-guided heterogeneous graph collaborative embedding for circRNA–drug sensitivity association prediction\\
\\
The program is designed for circRNA–drug sensitivity association prediction.\\
\\
chao cao](/content/explore/5dea18fa-6f28-4b04-a4fd-3b17d73321b7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7265648/tree/v1/index.html)

[Social Sciences\\
\\
23 \| Apr \| 2026\\
\\
Temporal Structure Perception in Tonal and Atonal Music: An Online Behavioural Study\\
\\
Exemplar audio stimuli, de-identified (participant IDs were hashed with a secured secret key) data, and relevant MATLAB and R code\\
\\
Seung-Goo Kim](/content/explore/19aa2102-6744-4b39-aaeb-7d495dca75ac?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9257419/tree/v1/index.html)

[Bioinformatics\\
\\
22 \| Apr \| 2026\\
\\
The Interrelationship between Preconception Folate Nutritional Intake and Child Genetic Liability in the Risk of Childhood Acute Lymphoblastic Leukemia\\
\\
This repository contains the code used for the analyses presented in our manuscript:\\
\\
"The Interrelationship between Preconception Folate Nutritional Intake and Child Genetic Liability in the Risk of Childhood Acute Lymphoblastic Leukemia"\\
\\
Currently under consideration for publication in Cancer Epidemiology, Biomarkers & Prevention.\\
\\
Yijin Xiang et al.](/content/explore/f51fad30-6c39-43ee-8bbb-93bf3b7c7569?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7823398/tree/v1/index.html)

published in [Cancer Epidemiology, Biomarkers & Prevention](/content/explore?query=Cancer%20Epidemiology%2C%20Biomarkers%20%26%20Prevention&refine=journal/index.html)

[Computer Science\\
\\
22 \| Apr \| 2026\\
\\
Beyond Uniform Forgiveness: Introducing the Selective Collar to Mitigate Evaluation Bias in Speaker Diarization\\
\\
Reference implementation and evaluation harness for the **Selective Collar**\\
diarization evaluation method, which conditionally snaps predicted speaker\\
boundaries to ground-truth boundaries when they fall within a small tolerance\\
window.\\
\\
Shah Manan Vinod](/content/explore/e0327a44-507e-478f-b1eb-2e6bfcce23ef?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6448410/tree/v1/index.html)

[Engineering\\
\\
22 \| Apr \| 2026\\
\\
Can industrial overcapacity enable seasonal flexibility in electricity use? A case study of aluminum smelting in China](/content/explore/b30127fb-a6bb-4bb6-9c17-56ca89e32676?page=2&filter=all/index.html)

[PyPSA-China is an open-source optimization model for the Chinese energy system built on the](/content/explore/b30127fb-a6bb-4bb6-9c17-56ca89e32676?page=2&filter=all/index.html) [PyPSA](https://pypsa.org/) framework. The model enables capacity expansion planning and operational optimization of China's power system, with special focus on aluminum smelting integration and grid flexibility.

Ruike Lyu

[Open Capsule](/content/capsule/0160517/tree/v1/index.html)

published in [Nature Energy](/content/explore?query=Nature%20Energy&refine=journal/index.html)

[Bioinformatics\\
\\
21 \| Apr \| 2026\\
\\
SpecGP: A Transformer-Based Model for Predicting Energy-adaptable Structural Spectra of Glycopeptides\\
\\
SpecGP is a transformer-based model for predicting energy-adaptable structural spectra of glycopeptides. This project was developed using Python 3.9.13. A GPU-enabled environment is strongly recommended for accelerated computation.\\
\\
Shisheng Sun](/content/explore/a7641f20-1684-455b-9513-72c86242ccb4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2279761/tree/v1/index.html)

published in [Nature Machine Intelligence](/content/explore?query=Nature%20Machine%20Intelligence&refine=journal/index.html)

[Biology\\
\\
4 \| May \| 2026\\
\\
Matlab scripts for: "Chemotactic self-organization captures the dynamics of mammalian hair follicle patterning "\\
\\
The spatial patterning of mammalian hair follicle precursors in embryonic skin is most commonly studied in the laboratory mouse (Mus musculus), where new follicles form equidistantly from pre-existing ones in successive rounds. This simple geometric rule has been effectively described as emerging from an expansion-induction process. However, such a description is incompatible with more recent developmental data indicating instead that scale, feather, and hair placodes self-organize through reaction–diffusion–chemotaxis cell interactions in the epidermis and dermis. Here, we suggest that the placode insertion mechanism observed in mammals is dominated by chemotactic self-organization. More specifically, we investigate a continuum dynamical model capturing interactions between dermal mesenchymal cells and an epidermal chemoattractant, embedded in a two-dimensional, isotropically expanding domain representing the growing embryonic skin. Using numerical simulations and comparison to experimental developmental data, we show that the chemotaxis model gives rise to the effective geometric rule that initially justified the development of the expansion-induction model in the laboratory mouse.\\
\\
Muhamet Ibrahimi, Ebrahim Jahanbakhsh](/content/explore/cfcd1786-3c66-4abb-898d-d8aaca62941d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5330829/tree/v2/index.html)

[Engineering\\
\\
20 \| Apr \| 2026\\
\\
A Decentralized Technique for Mean-Field Stochastic Delay Systems with Infinite Decision Makers and Related Remarks\\
\\
This capsule serves as supplementary material for the paper “A Decentralized Technique for Mean-Field Stochastic Delay Systems with Infinite Decision Makers and Related Remarks,” submitted to SICE FES 2026. It includes the implementation code, appendix, and full version of the paper. In particular, the capsule provides the computation of the exact optimal solution set for a finite number of decision-makers N, as well as an approximate optimal solution set for the infinite-population case. Furthermore, an error evaluation is conducted by substituting both solutions into the cost function.\\
\\
Zihang Tian](/content/explore/1a21fee0-546e-4a13-86f5-7c690510094d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9406355/tree/v1/index.html)

[Computer Science\\
\\
20 \| Apr \| 2026\\
\\
Hybridization of Lateralization and Landmark-based Approach to Emotion Categorization\\
\\
This repository includes all components of the proposed framework, including preprocessing, feature extraction, model training scripts, evaluation pipelines, and all attacks launched to the images as part of the study.\\
All code required to reproduce the results reported in the paper is provided, including both training and testing workflows. The corresponding datasets are not included and must be obtained directly from the respective data custodians in accordance with their usage policies.\\
\\
Harisu Abdullahi Shehu, Will Browne & Hedwig Eisenbarth](/content/explore/b9324536-74e4-4612-9995-41944b47493d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4420747/tree/v1/index.html)

[Computer Science\\
\\
20 \| Apr \| 2026\\
\\
Intelligent Systems for Academic Research Integration (ISARI): A Local and Fully Offline Brainstorming Partner for Ethical Scholarly Inquiry\\
\\
Artificial intelligence in education (AIEd) has rapidly expanded through applications centered on tutoring, feedback, personalization, writing assistance, and human–AI collaboration. Recent work has begun to examine generative AI as a brainstorming partner, a writing assistant, and a complementary collaborator in educational contexts. Yet a major gap remains: the field has paid far less attention to the infrastructures required for scholars to think with AI while preserving data privacy, interpretive authority, and institutional trust. This position paper introduces Intelligent Systems for Academic Research Integration (ISARI) as a response to that gap. ISARI is positioned as a local, fully offline, open-source, and explicitly multimodal brainstorming partner designed to support scholarly inquiry rather than replace it. The paper advances three claims. First, current AIEd conversations overemphasize cloud-based tutoring and writing assistance while underexamining research-facing infrastructures. Second, privacy, data stewardship, and trust should be treated as core design variables rather than secondary compliance issues. Third, a local AI thinking partner can broaden access to data science for qualitative and mixed-methods researchers while also helping quantitative scholars better understand how qualitative evidence can be rigorously analyzed and integrated into research. Building on adjacent work on human–AI brainstorming, AI-assisted writing, complementary human–AI roles, and offline-first educational architectures, this paper argues that the next step in AIEd should include tools that support interpretation, memoing, comparison, synthesis, and evidence-grounded writing without requiring the upload of sensitive materials to external servers. The revised argument extends that claim further by showing why local scholarly infrastructures must now also support image-based and visualization-based outputs—such as charts, diagrams, scanned pages, screenshots, temporal visualizations, network graphics, and geo-contextualized displays—if they are to remain methodologically useful for contemporary research practice.\\
\\
Manuel S. Gonzalez Canche](/content/explore/a45abe6d-315f-423d-9276-3acb3aec57be?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8887972/tree/v1/index.html)

[Computer Science\\
\\
16 \| Apr \| 2026\\
\\
Dual Representation-based Light Field View Synthesis using Deformable Convolutional and Deep Residual Channel Attention Networks\\
\\
Light Field (LF) cameras simultaneously capture both intensity values and directional information of light rays in a single exposure, providing a unique perspective for computational photography and 3D geometry perception. However, existing LF cameras are constrained by sensor resolution, limiting their ability to capture high spatial and angular resolutions simultaneously. To mitigate these issues, various learning-based methods have been proposed to increase the angular resolution of captured LFs, known as LF View Synthesis (LFVS). Many of these methods either neglect essential geometric cues or rely on neural networks with large receptive fields, which restrict their ability to accurately exploit LF structural characteristics. In response to these challenges, this paper introduces a dual representation-based LFVS method that employs deformable convolutional and Deep Residual Channel Attention (DRCA) networks. The proposed method includes two main modules: (i) Coarse Light Field View Synthesis (CLFVS) for initial LFVS, and (ii) Coarse-To-Fine Refinement (CFR) for final quality enhancement. The CLFVS module relies on deformable convolutions to adaptively extract LF features using two parallel networks: (i) Spatial Feature Extraction (SPFE) network using a depth-dependent LFVS approach, and (ii) Angular Feature Extraction (AFE) network using a non-depth-dependent LFVS approach. The CFR module refines the CLFVS output using a DRCA network, which employs dense residual connections between residual groups instead of conventional convolutional layers. The DRCA network employs Residual Channel Attention Blocks (RCABs) to model inter-channel dependencies, selectively enhancing meaningful features while suppressing irrelevant ones. The proposed method achieves state-of-the-art performance on synthetic and real-world LF benchmarks.\\
\\
Muhammad Zubair, Paulo Nunes, Caroline Conti & Luís Ducla Soares](/content/explore/e026f83e-0676-41c8-9fa7-a159b1b3e2ad?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2077922/tree/v1/index.html)

[Social Sciences\\
\\
15 \| Apr \| 2026\\
\\
Rethinking the double empathy problem - Modeling how autistic and non-autistic groups learn about their own and each other \[Experiment 3 Model Free Analyses\]\\
\\
This capsule includes all model-free analyses presented in Experiment 3.\\
\\
Shannon Cahalan et al.](/content/explore/26641fcf-7d22-4f10-8f85-1647297addd0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6349722/tree/v1/index.html)

published in [Nature Mental Health](/content/explore?query=Nature%20Mental%20Health&refine=journal/index.html)

[Social Sciences\\
\\
15 \| Apr \| 2026\\
\\
Rethinking the double empathy problem - Modeling how autistic and non-autistic groups learn about their own and each other \[Experiment 2 Model-Free Analyses\]\\
\\
This capsule includes all model-free analyses and data presented in Experiment 2.\\
\\
Shannon Cahalan et al.](/content/explore/c1343381-ff41-4336-929d-cee454170759?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4043096/tree/v1/index.html)

published in [Nature Mental Health](/content/explore?query=Nature%20Mental%20Health&refine=journal/index.html)

[Social Sciences\\
\\
15 \| Apr \| 2026\\
\\
Rethinking the double empathy problem - Modeling how autistic and non-autistic groups learn about their own and each other \[Experiment 1 self-preference distribution comparisons\]\\
\\
This capsule builds distributions for each of the three groups self-preferences for the Experiment 1 analysis. The script then implements one and two sample KS tests and Levene's Test for comparison.\\
\\
Shannon Cahalan et al.](/content/explore/4feb5c4b-0786-4ad1-abe2-9758c627b3c2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9405906/tree/v1/index.html)

published in [Nature Mental Health](/content/explore?query=Nature%20Mental%20Health&refine=journal/index.html)

[Social Sciences\\
\\
15 \| Apr \| 2026\\
\\
Rethinking the double empathy problem - Modeling how autistic and non-autistic groups learn about their own and each other \[Experiment 1 Self-preference comparisons and relationships\]\\
\\
This capsule runs comparisons of aggregated self-preferences and item-level relationships by group as referred to in Experiment 1.\\
\\
Shannon Cahalan et al.](/content/explore/583b54f0-09cc-4baf-8ef8-74d1ba34ca5f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7322867/tree/v1/index.html)

published in [Nature Mental Health](/content/explore?query=Nature%20Mental%20Health&refine=journal/index.html)

[Engineering\\
\\
1 \| Jun \| 2026\\
\\
Staged GT3 Setup Optimization with Setup Conditioned Telemetry Response Modeling in Simulation\\
\\
Optimizing a high fidelity GT3 race car setup is a high dimensional, nonlinear problem in which small changes to mechanical parameters can affect lap time, handling balance, and vehicle stability. Existing motorsport AI studies largely emphasize racing line optimiza-tion, autonomous control, race strategy, or offline vehicle dynamics estimation, while the mechanical setup layer is often treated as fixed or tuned manually. This paper presents a staged simulator based setup optimization framework augmented with setup conditioned telemetry response modeling. A BMW Z4 GT3 was evaluated at the Red Bull Ring in As-setto Corsa under a fixed AI driving policy across 134 setup configurations. The staged search improved best lap time from 91.430 s to 91.040 s, corresponding to a 0.390 s reduc-tion. To move beyond a single aggregate lap time claim, the full telemetry corpus was pro-cessed into 585 stable laps and 29,250 track position segment samples. A setup condi-tioned LightGBM model was trained to predict segment time and local vehicle response metrics from setup parameters and segment context, using five fold GroupKFold valida-tion by telemetry file to avoid random row leakage. The setup conditioned segment model reconstructed held out file level lap time with 0.223 s mean absolute error and Spearman correlation of 0.961, outperforming a setup only model at 0.288 s, a track only segment model at 0.687 s, and a shuffled setup placebo at 0.776 s. The same setup conditioned model also improved prediction of segment level speed, slip angle, tire load spread, rake, tire temperature, yaw response, and lateral acceleration. These results show that high frequency telemetry can support not only staged setup search, but also quantifiable learning of where and how setup changes alter vehicle behavior around the lap.\\
\\
Shanmukha Srivathsav Satujoda](/content/explore/4aa1cb8c-11d5-4ca8-aaa9-4df7da946712?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1977585/tree/v2/index.html)

[Engineering\\
\\
14 \| Apr \| 2026\\
\\
Assessment and optimisation of regional scale wind farm deployment using machine learning\\
\\
Code reproducing a minimal example of the workflow from "Assessment and optimisation of regional scale wind farm deployment using machine learning"\\
\\
Simon C. Warder, Mariana C. A. Clare, B. Bhaskaran & Matthew D. Piggott](/content/explore/99a1a20f-c2da-4bff-b15e-181529dc0d7f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7712896/tree/v1/index.html)

[Bioinformatics\\
\\
14 \| Apr \| 2026\\
\\
StepFold: A Progressive Local-to-Global Generation Framework for RNA Secondary Structure Prediction](/content/explore/4db5b55a-85e3-488b-a7c2-867247f863a6?page=2&filter=all/index.html)

[StepFold is a novel, multi-step generation framework designed for accurate and efficient RNA secondary structure prediction. Unlike conventional "one-step" deep learning methods, StepFold mimics the hierarchical nature of RNA folding by predicting structures progressively—from local substructures to global long-range interactions.\\
\\
This capsule enables rapid reproduction of paper results and allows users to perform RNA secondary structure prediction on new sequences using the pre-trained parameters.](/content/explore/4db5b55a-85e3-488b-a7c2-867247f863a6?page=2&filter=all/index.html)

[Official code is also available on GitHub:](/content/explore/4db5b55a-85e3-488b-a7c2-867247f863a6?page=2&filter=all/index.html) [https://github.com/ChengWang-hit/StepFold.git](https://github.com/ChengWang-hit/StepFold.git)

* * *

### 1\. Reproducing the Paper's Results

To reproduce the results presented in the paper, please follow these steps:

1. Open the `/code/run.sh` file.
2. Insert the following command into the script:

bash

Copy

```bash
python /code/StepFold-main/code/test_all.py

3. Click the Reproducible Run button.
4. Once the run is complete, the evaluation metrics will be saved and available in the /results/output.txt file.

2. Predicting Your RNA Sequences

To use StepFold for predicting the secondary structure of your RNA sequences, please follow these steps:

1. Open the /code/run.sh file.
2. Replace its content with the following command:

bash

Copy

python /code/StepFold-main/code/inference_fasta.py

3. Click the Reproducible Run button.
4. The prediction results will be saved to the /results directory. For each RNA sequence, the output will include the following files:
   • .bpseq file: The predicted secondary structure in BPSEQ format.
   • Arc diagram: A visualization of the secondary structure (.png).
   • Probability matrix: The base-pairing probability matrix (.npy).
   • Probability heatmap: A visual heatmap of the pairing probabilities (.png).
   • Contact map: The binary contact map array (.npy).

Cheng Wang et al.

Open Capsule

Biology\ \ 13 | Apr | 2026\ \ Comparative ASCL1 interactome analysis reveals CDK2-Cyclin A2 as suppressors of differentiation in MYCN-amplified neuroblastoma\ \ Scripts used to perform the analysis described in "Comparative ASCL1 interactome analysis reveals CDK2-Cyclin A2 as suppressors of differentiation in MYCN-amplified neuroblastoma"\ Mykhaylechko et al 2025\ \ Lidiya Mykhaylechko & Laura Woods Open Capsule

Engineering\ \ 13 | Apr | 2026\ \ Control occupation kernel regression for nonlinear control-affine systems\ \ This manuscript presents an algorithm for obtaining an approximation of a nonlinear high order control affine dynamical system. Controlled trajectories of the system are leveraged as the central unit of information via embedding them in vector-valued reproducing kernel Hilbert space (vvRKHS). The trajectories are embedded as the so-called higher order control occupation kernels which represent an operator on the vvRKHS corresponding to iterated integration after multiplication by a given controller. The solution to the system identification problem is then the unique solution of an infinite dimensional regularized regression problem. The representer theorem is then used to express the solution as finite linear combination of these occupation kernels, which converts an infinite dimensional optimization problem to a finite dimensional optimization problem. The vector valued structure of the Hilbert space allows for simultaneous approximation of the drift and control effectiveness components of the control affine system. Several experiments are performed to demonstrate the effectiveness of the developed approach.\ \ Moad Abudia, Tejasvi Channagiri, Joel A. Rosenfeld & Rushikesh Kamalapurkar Open Capsule

Associated article published in IEEE Transactions on Automatic Control

Engineering\ \ 13 | Apr | 2026\ \ Spectrum-Aligned FFT for THz Massive MIMO: A Pure Mathematical Solution to the Bandwidth Disaster\ \ Code for Spectrum-Aligned FFT for THz Massive MIMO: A Pure Mathematical Solution to the Bandwidth Disaster\ \ He Liao Open Capsule

Engineering\ \ 13 | Apr | 2026\ \ Task-adaptive eigenvector-based projection (EBP) transform for compressed sensing\ \ The compressed sensing (CS) theory requires the signal to be sparse under some\ transform. For most signals (e.g., speech and photos), the non-adaptive transform\ bases, such as discrete cosine transform (DCT), discrete Fourier transform (DFT), and\ Walsh-Hadamard transform (WHT), can meet this requirement and perform quite\ well. However, one limitation of these non-adaptive transforms is that we cannot\ leverage domain-specific knowledge to improve CS efficiency. This study presents a\ task-adaptive eigenvector-based projection (EBP) transform. The EBP basis has an\ equivalent effect of the principal component loading matrix and can generate a sparse\ representation in the latent space. In a Raman spectroscopic profiling case study, EBP\ demonstrates better performance than its non-adaptive counterparts. At the 1% CS\ sampling ratio (k), the reconstruction relative mean square errors of DCT, DFT, WHT\ and EBP are 0.33, 0.68, 0.32, and 0.00, respectively. At a fixed k, EBP achieves much\ better reconstruction quality than the non-adaptive counterparts. For specific domain\ tasks, EBP can significantly lower the CS sampling ratio and reduce the overall measurement cost.\ \ Yinsheng Zhang Open Capsule

Associated article published in Analytical Science Advances

Engineering\ \ 10 | Apr | 2026\ \ Tree Search Algorithms Applied to the BD-RIS Configuration in MU-MISO Communication Systems\ \ The reconfigurable intelligent surface (RIS) has attracted considerable attention of both academia and industry in recent years, given its capacity to dynamically manipulate the reflection of incident electromagnetic waves. Although the research developed for the RIS may have reached its maturity, there are still contentious aspects and limitations regarding its potential benefits for the next generation of wireless communications. In order to improve upon the the RIS technology, the beyond diagonal reconfigurable intelligent surface (BD-RIS) was recently proposed as an promising alternative. The BD-RIS boasts a more sophisticated circuit topology that is capable of providing more combinations of different adjustments or configurations for signal reflection. However, to aptly reap the benefits of the BD-RIS, the added degrees-of-freedom of its configuration must be leveraged accordingly. Therefore, in this work we propose a depth-first tree search algorithm for configuring the BD-RIS in multi-user multiple-input single-output (MU-MISO) communication systems. Taking advantage of the tree search exploration, the proposed algorithm achieves a remarkable trade-off between channel strength maximization performance and computational complexity scalability.\ \ Pedro H. C. de Souza & Luciano Mendes Open Capsule

Computer Science\ \ 10 | Apr | 2026\ \ An Adaptive Large Neighborhood Search for the Multiple Traveling Salesman Problem with Backup Coverage\ \ Adaptive Large Neighborhood Search meta heuristic solver for the multiple traveling salesman problem with backup coversge problem.\ \ Jonathan Cardozo Maciel, Guilherme Dhein & Olinto César Bassi de Araújo Open Capsule

Computer Science\ \ 10 | Apr | 2026\ \ A Network Traffic Matrix Estimation Method Based on Improved Multi-Granularity OD Pair Hierarchical Partitioning\ \ This code is about estimating the traffic matrix.\ \ Wenyue Sun Open Capsule

Bioinformatics\ \ 9 | Apr | 2026\ \ Single-cell insights into TFAM-mediated immunoregulation in dendritic cells of residual esophageal squamous cell carcinoma after neoadjuvant immunochemotherapy\ \ Codes for Single-cell insights into TFAM-mediated immunoregulation in dendritic cells of residual esophageal squamous cell carcinoma after neoadjuvant immunochemotherapy\ \ Linyan Chen Open Capsule

Engineering\ \ 9 | Apr | 2026\ \ DDoSimu5G: A simulation framework for generating labeled DDoS traffic datasets in 5G network\ \ DDoSimu5G v2.0 is a simulation framework for generating labeled DDoS traffic datasets in 5G network environments. It extends OMNeT++, INET, and Simu5G with application-layer modules for concurrent benign and adversarial traffic generation using configurable protocol profiles. Supports four DDoS attack models (UDP flood, TCP SYN flood, DNS amplification, HTTP flood) with dual-channel ground-truth labeling (in-band IPv4 TOS markers and out-of-band CSV annotations).\ \ Karim Khalil Open Capsule

Chemistry\ \ 9 | Apr | 2026\ \ Platonic representation of foundation machine learning interatomic potentials\ \ Code for extracting embeddings; anchor selection; embeddings transformation; optimal transport calculation; embedding similarity check; and equivariance perservation of materials with symmetrical atoms.\ \ Zhenzhu Li Open Capsule

published in Nature Machine Intelligence

Bioinformatics\ \ 9 | Apr | 2026\ \ pUniFind: A Unified, Large-Scale Pretrained Model for Pushing the Limits of Mass Spectra Interpretation\ \ This is the official repository for pUniFind, the most powerful zero-shot open peptide-spectrum scoring model surpassing other SOTA search engines and the first zero-shot open de novo sequencing deep learning model supporting over 1300 modifications. Developed by pFind group and DP Technology.\ \ Jiale Zhao Open Capsule

published in Nature Machine Intelligence

Bioinformatics\ \ 8 | Apr | 2026\ \ Two-Dimensional Geometric Template Diffusion for Boosting Single-Sequence Protein Structure Prediction\ \ The source code of a novel single-sequence protein structure prediction method named TDFold.\ \ Xudong Wang Open Capsule

published in Nature Machine Intelligence

Physics\ \ 8 | Apr | 2026\ \ AlphaXtal: Crystal Structure Prediction Towards Ground-Truth Structures via Monte Carlo Tree Search on a Discrete Mesh Grid\ \ AlphaXtal is a highly efficient Crystal Structure Prediction (CSP) program. By discretizing the infinite, continuous potential energy surface (PES) into a finite space and combining Monte Carlo Tree Search (MCTS) with Machine Learning Potentials (MLP), AlphaXtal enables high-throughput, systematic exploration of global minima and metastable structures within complex chemical spaces.\ \ MaoyanZhang Open Capsule

Engineering\ \ 8 | Apr | 2026\ \ Integral passivity-based control of underactuated mechanical systems with hysteretic elasticity\ \ README\ This folder contains supplementary files for the IEEE CDC 2024 submission \ "Integral passivity-based control of underactuated mechanical systems with hysteretic elasticity" \ authored by Enrico Franco, Fahim Shakib, and Kaiwen Chen.\ \ The MATLAB code is provided as is. \ It can be freely modified and reused, provided that the authors have been acknowledged.\ \ HOW TO \ Save all files in the active MATLAB folder. \ \ Run "script_BoB_Fig2" to call the function "odeBoB_Fig2" corresponding to Figure 2.\ \ Run "script_BoB_Fig3" to call the function "odeBoB_Fig3" corresponding to Figure 3.\ \ Run "script_BoB_Fig4" to call the function "odeBoB_Fig4" corresponding to Figure 4.\ \ NOTE\ The code produces the results for each setting separately (i.e., \alpha = 0.2, \alpha = 0.6, \alpha = 0.9 in Figure 2 and 4; s = 1, s = 2, s = 3 in Figure 3).\ Instead, Figure 2, 3, and 4 aggregate all results (i.e., for \alpha or for s).\ \ Enrico Franco, Fahim Shakib & Kaiwen Chen Open Capsule

Computer Science\ \ 8 | Apr | 2026\ \ FBNO Deep Neural Operator for Free Boundary Problems\ \ Free boundary problems (FBPs), characterized by partial differential equations defined on a priori unknown domains, arise across diverse scientific and engineering disciplines, from quantum physics to biomedical applications. Traditional numerical approaches face significant limitations. While recent advances in neural operators have revolutionized PDE solving by learning mappings between function spaces, existing frameworks remain constrained to predefined domains, rendering them inapplicable to FBPs. Here, we introduce the Free Boundary Neural Operator (FBNO), a universal framework that overcomes this fundamental limitation by leveraging topological conjugacy between dynamical systems. FBNO approximates both the flow map of a conjugate system and the homeomorphism linking it to the original FBP, enabling predictions on evolving domains without prior geometric knowledge. Crucially, we provide an approximation theorem guaranteeing the method’s theoretical feasibility. The efficacy of FBNO has been comprehensively demonstrated in numerical experiments spanning phase transitions, non-convex geometries, and multi-physics systems. The work marks a turning point in free boundary simulations, where neural networks unlock both speed and precision.\ \ Zongjia Long Open Capsule

published in Nature Machine Intelligence

Physics\ \ 8 | Apr | 2026\ \ Beyond the Electric Dipole Approximation: Electric and Magnetic Multipole Contributions Reveal Biaxial Water Structure from SFG Spectra at the Air-Water Interface\ \ A program that computes multipolar SFG spectra.\ \ Louis Lehmann et al. Open Capsule

published in Nature Communications

Engineering\ \ 7 | Apr | 2026\ \ FLOPpy: A hardware-agnostic Python library to monitor the computational cost of Machine and Deep Learning algorithms\ \ FLOPpy is a versatile Python library designed to monitor and estimate the algorithmic workload of both Deep Learning (PyTorch) and Machine Learning (Scikit-learn) models.\ \ By systematically tracking Floating Point Operations (FLOPs) and BOPs (Bit-OPerations), it provides a hardware-independent assessment of the total computational demand, spanning from standard Forward and Backward passes to Optimizer updates and Loss evaluations.\ \ Francesco Scala, Francesco Mandarino, Liliana Martirano & Luigi Pontieri Open Capsule

Computer Science\ \ 7 | Apr | 2026\ \ Imaging Hidden Objects with Consumer LiDAR using Motion-Induced Aperture Sampling (round 2))\ \ Code to image hidden objects from consumer LiDAR. Vision capabilities include object tracking, camera localization, and 3D reconstruction.\ \ Siddharth Somasundaram Open Capsule

published in Nature

Biology\ \ 7 | Apr | 2026\ \ Predicting temporal stability and resilience from resistance and recovery\ \ This capsule contains all R code and data used to reproduce the analyses, figures, and tables for the paper "Predicting temporal stability and resilience from resistance and recovery." All results can be reproduced by running the stabilityCode script.\ \ Forest Isbell Open Capsule

published in Nature

Engineering\ \ 7 | Apr | 2026\ \ Eclipsing Tolerant Codes\ \ Matlab scripts used in the publication of the Eclipsing Tolerant Codes paper.\ \ Matthew Bayer & Adly Fam Open Capsule

Engineering\ \ 18 | May | 2026\ \ Mechanism Design and Co-operative Game Theory for Cybersecurity Defenders - One Combined Game\ \ A Game Theoretic plugin for Security Engineers, to help developers write secure code. The plugin extends functionality of a security code review tool called Bandit. See this plugin's readme for more details.\ \ v1.0 was the main release\ The next version implemented error handling for a corner case involving testing of 1 test file (hello.py)\ \ Mithun Vaidhyanathan, Weisheng Si, Bahman Javadi & Seyit Camtepe Open Capsule

Computer Science\ \ 7 | Apr | 2026\ \ Matching Theory-Based Resource Allocation and Spectrum Sharing for UAV–RIS Assisted Cellular–IoT Systems\ \ Unmanned aerial vehicles (UAVs) and reconfigurable intelligent surfaces (RISs) are two emerging technologies\ envisioned for sixth-generation (6G) wireless systems. These\ technologies enhance conventional cellular networks by extending\ coverage and enabling ubiquitous connectivity. However, spectrum scarcity and interoperability challenges create a strong need for efficient spectrum sharing among cellular users supported\ by these technologies. In this context, this paper considers a\ dynamic spectrum-sharing method in which data rate–aware\ spectrum sharing plays a critical role in managing base station\ power consumption and mitigating interference. We investigate\ a UAV–RIS-assisted cellular system where legacy cellular users\ share the spectrum with cellular Internet-of-Things (IoT) devices.\ To enhance the overall system sum data rate, we propose a joint\ user pairing, spectrum, power allocation, and RIS phase shift\ optimization approach based on matching theory. Simulation\ results demonstrate the effectiveness of the proposed method in\ improving resource allocation efficiency and significantly enhanc-\ ing the sum data rate performance of wireless communication\ systems.\ \ Lilatul Ferdouse Open Capsule

Medical Sciences\ \ 7 | Apr | 2026\ \ ExtractSUS - Open source for automated data extraction from DataSUS

Automated data download from DataSUS of the Brazilian Unified Health System (SUS - Sistema Único de Saúde). Notebook with open-source code for downloading data from the DataSUS Health Information Systems (SIS - Sistemas de Informação em Saúde) in a simple and intuitive way.

https://github.com/zehgobbes/ExtractSUS

Available systems:

• SIM - Mortality Information System;
• SINASC - Live Birth Information System;
• SINAN - Notifiable Diseases Information System;
• SIA - Outpatient Information System;
• SIH - Hospital Information System.

José Alberto Gobbes Cararo, Danyelle Oliveira Fonte & Ana Laura Sene Amâncio Zara

Open Capsule

Earth Sciences\ \ 3 | Apr | 2026\ \ The Code of Freeze-Thaw-Driven Soil Moisture Return Dominates Spring Phenology on the Warming Qinghai-Tibet Plateau\ \ This is the code for the study of Freeze-Thaw-Driven Soil Moisture Return Dominates Spring Phenology on the Warming Qinghai-Tibet Plateau.\ \ Hanrui Zhao et al. Open Capsule

published in Nature Communications

Earth Sciences\ \ 2 | Apr | 2026\ \ A generic deep learning model for global nearshore wave assessment\ \ This capsule contains the source code and data for SWAN-TC, a physics-guided hybrid framework for high-resolution nearshore wave prediction. It includes the full implementation of the SWAN-T (Transformer surrogate), SWAN-C (residual correction), and the integrated SWAN-TC models.\ \ Key Contents:\ \ 1、Model Code.\ \ 2、Visualization Data: Processed datasets required to generate the figures.\ \ 3、Reproducibility: Clicking "Reproducible Run" will execute the main script to generate the figures presented in the main text of the manuscript.\ \ Xi Lin Open Capsule

Computer Science\ \ 2 | Apr | 2026\ \ Towards responsible AI for plague foci surveillance and outbreak prediction: Mongolian case\ \ This capsule provides a machine learning framework to identify and classify active vs. inactive plague focal points using historical data (2015-2025). The core model is built using the XGBoost algorithm to predict high-risk areas based on environmental and epidemiological factors.\ \ Tuyatsetseg Badarch Open Capsule

Computer Science\ \ 2 | Apr | 2026\ \ Code for "Depth Enhanced Cascaded Framework for OCTA Segmentation with Structureand Connectivity-Aware Losses"\ \ Optical coherence tomography angiography (OCTA), known for its high-resolution and noninvasive imaging capability, has become a key modality for visualizing retinal vasculature. Accurate and automated segmentation of capillaries, arteries, veins, and foveal avascular zone in OCTA images is essential for quantitative analysis and disease assessment. In this paper, we propose a depth enhanced cascaded framework specifically designed for multi-class OCTA segmentation. Our method investigates the spatial distribution of vasculature in retinal images and integrates a novel self-supervised depth prediction module to learn implicit depth cues from volumetric data, thereby improving the discrimination of overlapping vascular layers. In addition, we design two topology-aware loss functions that explicitly encourage structural integrity and continuity of vessel segmentation, particularly at bifurcations and endpoints. Experiments on the OCTA-6mm and OCTA-3mm datasets demonstrate that our method outperforms existing state-of-the-art approaches, with mIoU gains of around 2% over prior method, IPNv2, thereby highlighting enhanced segmentation accuracy and vascular topology preservation.\ \ Data Availability Statement:\ The dataset used in this capsule is a public dataset obtained from OCTA-500. All rights to the dataset belong to the original authors. If you use this dataset, please cite their original work: \ @article{li2024octa,\ title={OCTA-500: a retinal dataset for optical coherence tomography angiography study},\ author={Li, Mingchao and Huang, Kun and Xu, Qiuzhuo and Yang, Jiadong and Zhang, Yuhan and Ji, Zexuan and Xie, Keren and Yuan, Songtao and Liu, Qinghuai and Chen, Qiang},\ journal={Medical image analysis},\ volume={93},\ pages={103092},\ year={2024},\ publisher={Elsevier}\ }.\ \ Bisheng Wang, Yuexuan Wang, Jaime S. Cardoso & Lin Wu Open Capsule

Chemistry\ \ 1 | Apr | 2026\ \ Autonomous Interface Engineering for Highly Efficient and Reproducible Perovskite Solar Cells\ \ python script for the design of passivation molecular for perovskite solar cell.\ \ Shuaihua Lu Open Capsule

published in Nature

Computer Science\ \ 1 | Apr | 2026\ \ Evaluation Code for SABLE: A Sparse Bayesian Meta-Learning Framework for Imbalanced Few-Shot Traffic Anomaly Detection\ \ The evaluation code for the paper "SABLE: A Sparse Bayesian Meta-Learning Framework for Imbalanced Few-Shot Traffic Anomaly Detection"\ \ Hujite Lin Open Capsule

Bioinformatics\ \ 1 | Apr | 2026\ \ Uncovering Non-apparent Visual Encodings in Latent Space (UNVEIL)\ \ This repository implements UNVEIL, a framework for identifying, quantifying, and mitigating demographic-associated signals in pathology foundation model representations. By integrating demographic classification, nuclear morphometric analysis, and demographic signal-aware agentic scheduling, UNVEIL addresses performance disparities in computational pathology tasks.\ \ Ting-Wan Kao Open Capsule

Engineering\ \ 1 | Apr | 2026\ \ ConsBASE: End-to-End ETL Pipeline for Korean Construction Document Integration\ \ ConsBASE is an end-to-end document analysis system designed to convert scanned construction documents into structured JSON outputs.\ The framework integrates OCR, table detection, image detection, and text structuring into a unified ETL pipeline.\ \ Seung-Kyu Hong & Dong-Seok Seo Open Capsule

Bioinformatics\ \ 1 | Apr | 2026\ \ A stochastic threshold model of LRP6 expression underlies incomplete penetrance in human tooth agenesis\ \ Heterozygous loss-of-function variants in LRP6, encoding a Wnt co-receptor essential for tooth development, cause autosomal dominant oligodontia with incomplete penetrance. Few quantitative frameworks have addressed why carriers of identical pathogenic variants present with different dental phenotypes. Whole-exome sequencing in a Thai family segregating oligodontia identified a novel canonical splice-site variant, NM_002336.3(LRP6):c.845-1G>A. RT-PCR confirmed complete exon 5 skipping, producing an in-frame deletion of 44 amino acids (p.Val282_Gln325del) within the second beta-propeller domain at the Wnt ligand-binding interface. The variant was classified as likely pathogenic (PVS1_moderate, PM2, PP1, PP3, PP4). Single-cell RNA-seq of mouse molar development (74,107 cells) revealed Lrp6 expression in only 36.5% of dental mesenchyme cells, with high cell-to-cell variability (coefficient of variation = 0.352, 2.5-fold higher than Actb). Cells with higher Lrp6 expression showed significant but negligible association with Wnt pathway activity (Kruskal-Wallis H = 110.28, p = 1.13 × 10 -24 ; eta-squared = 0.003), consistent with LRP6 functioning as a permissive threshold gatekeeper rather than a linear driver. Monte Carlo simulations demonstrated that under haploinsufficiency, incomplete penetrance naturally emerges from stochastic expression heterogeneity: at permissive parameters, 86% of heterozygous tooth germs developed normally, yielding 14% disease penetrance. In human adult dental tissue, LRP6 was detected in 15.8% of mesenchymal cells, indicating that restricted expression persists beyond embryonic development. While this adult dataset does not correspond to the mouse embryonic stage, it suggests low LRP6 prevalence is a sustained dental mesenchyme lineage feature, not a transient artifact.\ \ Tohid Ghasemnejad et al. Open Capsule

Computer Science\ \ 1 | Apr | 2026\ \ Bridging the Interpretability Gap for Medical AI Models using Class-Association Manifold Learning\ \ We provide related codes and data of the Class-Associated-Manifold Learning (CAML) here. This project includes the training of the CAML model, the analysis of the class-associated codes, and the case show of the instance explanation.\ \ Ruitao Xie et al. Open Capsule

published in Nature Biomedical Engineering

Physics\ \ 1 | Apr | 2026\ \ Field-resolved Observation of Exciton Coherence in a van der Waals Magnet\ \ STFT code + data files\ \ Matthew Yeung Open Capsule

published in Nature Materials

Computer Science\ \ 31 | Mar | 2026\ \ Learning Network Dismantling Without Handcrafted Inputs

Demo for MIND, an RL-driven network dismantler that finds the sequence of node removals that most rapidly fragments a network into disconnected components.

Learning Network Dismantling Without Handcrafted Inputs\ \ Haozhe Tian, Pietro Ferraro, Robert Shorten, Mahdi Jalili, Homayoun Hamedmoghadam\ \ AAAI-26 Main Technical Track (Oral) [arXiv]

Haozhe Tian

Open Capsule

Associated article published in Proceedings of the AAAI Conference on Artificial Intelligence

Bioinformatics\ \ 31 | Mar | 2026\ \ FedDHS-Synth: A Federated Differentially Private Framework for Synthetic Vaccination Survey Data Generation in South Asia\ \ FedDHS-Synth is a federated differentially private framework that generates synthetic childhood vaccination survey data from Bangladesh and Pakistan DHS microdata without centralising sensitive records. Synthetic data retains 98.9% of downstream predictive utility while membership inference attacks confirm near-random privacy leakage (AUC 0.51), enabling privacy-preserving multi-country public health research in low-resource settings.\ \ Shahan Ahmed Open Capsule

Computer Science\ \ 31 | Mar | 2026\ \ RL-Driven Sequential Matched Filtering for ECG R-peak Detection

This capsule implements the Sequential Matched Filter (SMF) algorithm, a lightweight, interpretable ECG R-peak detector that uses matched filters (MFs) to iteratively refine signals.

Machine Intelligence on the Edge: Interpretable Cardiac Pattern Localisation Using Reinforcement Learning\ \ Haozhe Tian, Qiyu Rao, Nina Moutonnet, Pietro Ferraro, Danilo Mandic\ \ Machine Intelligence Research [arXiv]

Haozhe Tian & Qiyu Rao

Open Capsule

Bioinformatics\ \ 31 | Mar | 2026\ \ Dynamic Entropy Search Enables Scalable and Updatable Fast Spectral Comparison in Expanding Repositories\ \ Spectral entropy is a useful property to measure the complexity of a spectrum. It is inspired by the concept of Shannon entropy in information theory. \ Entropy similarity, which measured spectral similarity based on spectral entropy, has been shown to outperform dot product similarity in compound identification. \ The calculation of entropy similarity can be accelerated by using the Flash Entropy Search algorithm. \ Dynamic Entropy Search is built and optimized based on Flash Entropy Search. Besides the excellent search performance, it allows unlimited library spectra with high speed and low memory. Dynamic Entropy Search enables efficient index build, index update and library search of MS/MS spectra.\ \ Ren Guo Open Capsule

Physics\ \ 31 | Mar | 2026\ \ Charge-dependent spectral softenings of primary cosmic-rays from proton to iron below the knee\ \ Fitting for cosmic ray's spectrum with DAMPE's data.\ \ Chuan Yue Open Capsule

published in Nature

Computer Science\ \ 30 | Mar | 2026\ \ Robust Control Invariant Sets for Continuous, Fully Controllable Linear Systems using Zonotopes\ \ A robust control invariant set of a cyber-physical system describes a safe region of operation, for which we can guarantee at all times that the system can always be controlled in such a way that it never leaves this region. This concept has been studied in detail for discrete-time systems, leading to scalable approaches even for nonlinear systems. However, continuous-time systems have remained challenging due to the complexity of approximating the behavior of the system for very small times. In this study, we propose a novel approach based on zonotope containment and the analysis of the corresponding matrix norms, allowing us to tightly enclose all possible trajectories of the system at all times. We focus our attention on fully controllable linear systems, as a first step towards a more general framework.\ \ Adrian Kulmburg & André Platzer Open Capsule

Biology\ \ 30 | Mar | 2026\ \ Blood flow patterns are regulated by interpericyte tunneling nanotubes connecting functionally-opposite neuronal areas\ \ Blood flow patterns are regulated by interpericyte tunneling nanotubes connecting functionally-opposite neuronal areas\ \ Jesse Gardner-Russell et al. Open Capsule

published in Nature Communications

Engineering\ \ 27 | Mar | 2026\ \ KAN-Based Prediction and Optimization of Specific ON-resistance of VVD-SJ MOSFETs\ \ These files accompany the article “KAN-Based Prediction and Optimization of Specific ON-resistance of VVD-SJ MOSFETs,” to be published in IEEE Transactions on Electron Devices. The codes basisFunc_spline.m, modelKA_basisC.m, and splineMatrix.m are open-sourced by Michael Poluektov et al. (MIT); the license is included herein. BV_Ron_dataset.m is used for dataset generation, and KANmain_xxx.m is used for prediction result processing and evaluation.\ \ Zhiwei Jing & Haimeng Huang Open Capsule

Bioinformatics\ \ 26 | Mar | 2026\ \ Breaking the sparsity barrier GroupSig analysis\ \ Capsule Contents \ \ This capsule contains the code for running a genome-wide scan (GWAS) for Signature Quantitative Traits (SigQTLs) on genomic data. It comes with simulated data to run the analysis on, and can be easily run on real genomic data provided by the user by using raw vcf files (and using the raw vcf arg in the GroupSig_analysis - use python GroupSig_analysis .py -h to get all args) and a "mutational context" csv file which is a DataFrame generated as part of the output of the sigProfilerMatrixGenerator tool's SigProfilerMatrixGeneratorFunc function (grab it by using "your_variable_datafarme = output['96']"), which is applied to a MAF file after appropriate filtrations (we provide an example script). \ \ We also present the code for easily creating a graph of the linear regression in a specific variant under graph_snp.py which cannot be run on Code Ocean since it requires the results be generated first, but it cannot load them from the /results dir, and the code used to generate figures and the non-central analyses also presented in the article, for which data is not publicly available.\ \ Breaking the sparsity barrier abstract: \ \ Cancer development is shaped by somatic mutational processes that leave characteristic patterns known as mutational signatures. The inherited determinants of variability in signature activity remain largely unknown. Common germline variants that regulate this activity, which we term Signature Quantitative Trait Loci (SigQTLs), are expected to have modest individual effects, requiring cohorts of tens of thousands of samples for reliable detection. Clinical targeted-panel sequencing datasets achieve this scale, but present a fundamental challenge: individual tumors typically harbor too few mutations for stable signature inference. To overcome this sparsity barrier, we introduce GroupSig, a framework that aggregates sparse mutational patterns across samples sharing a germline genotype into information-rich meta-samples, enabling robust signature inference at the population level.\ We validated GroupSig by recovering the well-established correlations between age and clock-like signatures SBS1 and SBS5 using emulated panel data from The-Cancer-Genome-Atlas. We then applied GroupSig to approximately 32,000 tumor samples from the Dana-Farber Cancer Institute PROFILE cohort in a genome-wide SigQTL scan. We identified 9 genome-wide significant SigQTLs, with the strongest signal at locus 16q24.3, where six variants were associated with increased SBS7 (UV exposure) activity. This association persisted after excluding melanoma samples, arguing against a tumor-type enrichment artifact. Validation in TCGA confirmed 6 SigQTLs, all at 16q24.3, where implicated variants are eQTLs for CDK10 and SPG7 in skin tissue. Beyond genome-wide hits, DNA repair genes were 12.6-fold enriched among sub-threshold signals, supporting a polygenic architecture for mutational process regulation. GroupSig provides a scalable framework for germline-somatic association studies using panel sequencing data.\ \ Alon Ravid Open Capsule

Engineering\ \ 26 | Mar | 2026\ \ Pilot-Free Precoded OFDM with Spectrum-Aligned FFT Receiver for Extreme Doppler UWA Channels\ \ Pilot-Free Precoded OFDM with Spectrum-Aligned FFT Receiver for Extreme Doppler\ UWA Channels\ \ He Liao Open Capsule

Engineering\ \ 26 | Mar | 2026\ \ A Handheld Multispectral NIRS Device for Real-Time Baseline-Relative Monitoring of Tissue Oxygen Saturation - Models\ \ This Jupyter Notebook contains the Python code for the models used in the paper titled "A Handheld Multispectral NIRS Device for Real-Time Baseline-Relative Monitoring of Tissue Oxygen Saturation".\ \ Devang Vyas Open Capsule

Associated article published in IEEE Transactions on Biomedical Engineering

Mathematics\ \ 26 | Mar | 2026\ \ Dual IRLS scheme for graph p-Laplacians and l^p regression\ \ This MATLAB code accompanies the paper “The dual IRLS scheme for (hyper-)graph p-Laplacians and l^p regression with large exponents p”. It contains all scripts required to reproduce the figures in the manuscript. In particular, it solves a linear l^p regression problem (Experiment 1) and a classification problem based on the graph p-Laplacian with underlying MNIST (Experiment 2a) and fashion MNIST (Experiment 2b) data set.\ \ Johannes Storn Open Capsule

Physics\ \ 26 | Mar | 2026\ \ Observation of average topological phase in disordered Rydberg atom array\ \ Topological phases have been extensively studied over the past two decades, primarily in quantum pure states, where they are protected by exact symmetries. Recently, numerous studies have theoretically demonstrated the existence of average symmetry-protected topological (SPT) phases in mixed quantum states, which naturally arise in real systems due to decoherence or disorder. Despite extensive experimental observations of exact SPT phases in various systems, ranging from solid-state materials to synthetic matters, average SPT phases are yet to be observed until this work. Here we report direct observations of disorder-induced many-body interacting average SPT phase in an atom array at half-filling, whereby random offsets to tweezer locations forming a lattice implement structural disorder, resulting in fluctuating long-range dipolar interactions between tweezer confined single atoms. The induced topological phase is vindicated by the spatially resolved atom-atom correlation functions for different forms of dimer compositions. The ground state degeneracy in disordered configurations is detected and compared to the regular lattice without disorder. By probing the quench dynamics of a highly excited state, we observe markedly slower decay of edge spin magnetization in comparison to the bulk spin, consistent with the presence of topologically protected edge modes in disordered lattices.\ \ Zongpei Yue et al. Open Capsule

published in Nature Physics

[Computer Science\ \ 25 | Mar | 2026\ \ Efficient Unified Online Feature Selection via GMM Parameter Matrix and Binary Relevance-Redundancy Assessment\ \ This capsule provides the official Python implementation of UOFS (Efficient Unified Online Feature Selection via GMM Parameter Matrix and Binary Relevance-Redundancy Assessment).\ \ Feature selection is crucial for processing high-dimensional visual data and continuous image streams in modern computer vision and image processing tasks. However, many existing methods are limited to handling specific types of streams. UOFS brings a unified framework designed to efficiently handle varying dynamic data and feature streams by leveraging Gaussian Mixture Model (GMM) parameter matrices and a novel binary relevance-redundancy assessment strategy.\ \ Key Features & Capabilities:\ \

• **Versatile Stream Support:**The codebase inherently simulates and evaluates four different feature selection scenarios common in dynamic visual environments:\
  • Offline Feature Selection (offline)\
  • Data Stream Feature Selection (data)\
  • Feature Stream Feature Selection (feature)\
  • Trapezoidal Data Stream Feature Selection (trap)\
• Comprehensive Pipeline: The experiment runner handles everything from data scaling (StandardScaler) and UOFS execution to training a Support Vector Machine (Linear SVM) to measure the effectiveness of the selected features on image classification tasks.\
• Reproducibility: Includes multiple benchmark datasets (e.g., USPS, DrivFace, ORL—widely used in face recognition and digit classification) directly within the data/ directory.\ \ Click "Reproducible Run" to execute the pipeline. The console will display the selection process, the exact number of features retained, and the final classification performance (Accuracy & F1 Score).\ \ Junyang Wu, Linjing You & Jiabao Lu](/content/explore/6796cc53-27e8-4f38-890a-3645a3f07655?page=2&filter=all/index.html) Open Capsule

Physics\ \ 25 | Mar | 2026\ \ Pinhole Engineering based Enhanced Resolution (PEER) for Fluorescence Lifetime Imaging Microscopy\ \ This Python script streamlines the PEER workflow: it ingests six TIFF images—intensity (I) and phasor coordinates (G and S) captured with both the small‑pinhole and large‑pinhole settings of a confocal FLIM system—and automatically generates four figures: phasor‑density plots and intensity‑weighted lifetime maps for the confocal and PEER channels. The resulting PNG files — “phasor_confocal.png,” “phasor_peer.png,” “flim_confocal.png,” and “flim_peer.png” — are saved in the specified results folder.\ \ Wonsang Hwang et al. Open Capsule

Medical Sciences\ \ 24 | Mar | 2026\ \ Audiogram Detection AI Trained Using Size-Fixed Frame and Bounding-Boxes

[This capsule is provided to reproduce the results of the manuscript entitled\ "Audiogram Detection AI Trained Using Size-Fixed Frame and Bounding-Boxes" in New Generation Computing\ \ Datasets:\ data/datasets/training_datasets directory contains the following training/validation datasets:\ 1. DatasetA (rect416 images) for DETR or YOLO training\ 2. DatasetB (original images) for DETR training\ 3. DatasetC (original images) for YOLO training\ \ data/datasets/test_datasets directory contains the following test datasets (in YOLO format):\ 1. Original test dataset\ 2. Original test dataset without photo images\ 3. Rect416 test dataset\ \ Models:\ data/models/\ 1. rect416DETR_audiogramdetr_checkpoint0499.pth (trained with datasetA)\ classes: a0 (right air-conduction threshold), a1 (right bone-conduction threshold), \ a2 (left air-conduction threshold), a3 (left bone-conduction threshold), \ and a4 (overlapping right and left air-conduction thresholds).\ 2. rect416YOLO_audiogramdetr2yolo_best_20241023.pt (trained with datasetA)\ classes: the above a0-a4\ 3. originalDETR_checkpoint0499.pth (trained with datasetB)\ classes: aa0 (audiogram frame), aa1 (0-dB line),\ and aa2–aa6 (corresponding to a0–a4, respectively)\ 4. (conventional model)originalYOLO_aagraphmodel3_ntnshladded_aa6_best_20240315.pt (trained with datasetC)\ classes: the above aa0-aa6\ Utilities\ General: code/utilities/general\ 1. extract_resize2rect416image.py: to extract a rectangle from each original image and resize it to 416 x 416 pixcels\ 2. drawingfullgraph_rect416.py: to redraw a standardized audiogram for a rect416 format\ 3. confusionmatrix_testdataset.py: to depict a confusion matrix and calculate model performance\ \ Copy\ \

df_preps: code/utilities/df_preps\\
to prepare a DataFrame (df) from YOLOv5x detection results in respective Figures\\
```\\
\\
DETR: code/utilities/DETR/\\
1\. rect416DETRprediction1.py: to perform inference using rect416DETR model](/content/explore/ccceb860-f995-4363-b6d3-656f50d9e5f2?page=2&filter=all/index.html)

[YOLOv5\\
YOLOv5 is included in this capsule : License is the same as the YOLOv5 repository.](/content/explore/ccceb860-f995-4363-b6d3-656f50d9e5f2?page=2&filter=all/index.html) [https://github.com/ultralytics/yolov5](https://github.com/ultralytics/yolov5)
GNU Affero General Public License v3.0
DETR
DETR is included in this capsule : License is the same as the original DETR repository.
[https://github.com/facebookresearch/detr](https://github.com/facebookresearch/detr)
Apache License Version 2.0

Others:
The license used for each model is the same as the original license.

Tomoyuki Shishido et al.

[Open Capsule](/content/capsule/1236368/tree/v1/index.html)

[Computer Science\\
\\
24 \| Mar \| 2026\\
\\
Interval-Based Gas Emission Rate Estimation Using Single-Camera Occlusion Analysis\\
\\
The simulation provides controlled synthetic scenarios for validating the computational pipeline, including depth corridor construction, optical-flow-based kinematics, and interval propagation. Results include tables and plots demonstrating interval coverage, observability regimes, and degradation modes.\\
\\
Andrii Syrotenko & Azat Davliatshin](/content/explore/8c407ed6-f3a4-4570-af87-081f006ea1a4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8036568/tree/v1/index.html)

[Bioinformatics\\
\\
24 \| Mar \| 2026\\
\\
LABEL (Array-seq): Whole-body molecular and cellular mapping of the laboratory mouse\\
\\
**LABEL: Spatially Aware Histology-Based Classification** \\
\\
**LABEL** is a pipeline that integrates histology feature extraction with a spatially aware hierarchical classifier to predict **organ**, **tissue region**, and **cell type** identities from H&E-stained images.\\
\\
* * *\\
\\
**Pipeline overview**  **Preprocessing and feature extraction** \\
\\
H&E images are stain-normalized and segmented into image patches centered on spatial transcriptomic capture spots. Each patch is processed using a pretrained **UNI pathology foundation model** to generate fixed-dimensional feature embeddings.\\
\\
**Hierarchical prediction** \\
\\
Predictions are performed using a spatially aware **k-nearest neighbor (kNN)** classifier that incorporates image features and spatial coordinates. Independent classifiers are trained for each hierarchical annotation level:\\
\\
- Organ \\
- Tissue region\\
- Cell type\\
\\
**Training and evaluation schemes**\\
\\
- **In-distribution evaluation**\\
\\
Models are trained on a fixed subset of spatial spots and evaluated on held-out data from the same sections.\\
\\
- **Leave-one-section-out evaluation**\\
\\
Models are trained on multiple whole mouse-sections and evaluated on a fully held-out section.\\
\\
\\
**Performance assessment** \\
\\
Model performance is summarized using **normalized confusion matrices**, restricted to classes consistently observed across replicates.\\
\\
Bohan Li, Feng Bao, Margarette Clevenger & Nicolas Chevrier](/content/explore/df903e84-7768-47cf-9742-7f824af4a67a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7566764/tree/v1/index.html)

[Computer Science\\
\\
24 \| Mar \| 2026\\
\\
Speech-to-Text with Translation\\
\\
Speech-to-text tool with speaker diarization and optional (subsequent) translation of a transcript. The tool uses open-weights AI models.\\
\\
Andreas Lindner](/content/explore/89f80e25-2315-44c4-a205-314ba60198dc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9249326/tree/v1/index.html)

[Engineering\\
\\
24 \| Mar \| 2026\\
\\
Analytical Method for Calculation of Electromagnetic Fields from Armoured HVAC Cables\\
\\
With the proliferation of HVAC armoured cables in offshore wind applications, the provision of methods for the calculation of both magnetic and electric field effects from HV cables is of high importance for consenting processes associated with offshore wind development and biological experiments. Despite electro-reception of marine animals being a known phenomenon, the electric fields present outside of HVAC cables, mostly induced as a result of the time-varying magnetic vector potential, have not been thoroughly analysed in literature. The available methods for the calculation of magnetic fields are often complex and require an expert user. In this work, this knowledge gap is addressed through the development of an analytical method, with some parameters derived from simple numerical simulations, that can be applied to both magnetic and electric fields and by provision of MATLAB code for calculation of both effects. The presented method includes the impact of metallic sheaths, cable twist, and magnetic armour. For the development of electric field expressions, the impact of the insulating layer at the cable outer boundary is considered. The presented results are within 3% of the numerical solution for the magnetic and electric field norm and within 5% for the individual components of the fields.\\
\\
Joanna Rzempołuch et al.](/content/explore/506d367c-f395-409a-b258-c87044688bac?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4091046/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tpwrd.2026.3671454) published in [IEEE Transactions on Power Delivery](/content/explore?query=IEEE%20Transactions%20on%20Power%20Delivery&refine=journal/index.html)

[Engineering\\
\\
24 \| Mar \| 2026\\
\\
EFESOS: A GNN Framework for Early Fault-Propagation Estimation in SoCs for Functional Safety-Oriented Design\\
\\
Safety-critical applications increasingly challenge System-on-Chip (SoC) design due to rising complexity and computational demands. Meeting Functional Safety (FuSA) certification requires rigorous verification, often via tedious fault injection (FI) simulations to assess fault-propagation to system’s primary outputs, verify and identify vulnerabilities, and validate safety mechanisms. Recent evaluation methods utilize AI-based approaches to enhance fault-effect prediction by leveraging structural and workload characterizations. But these typically rely on late-stage design models. In this work, we propose EFESOS (Early Fault-Propagation Estimation in SoCs), a GNN-based framework for early estimation of fault-propagation criticality in SoCs using early RT-level models. Unlike existing approaches that rely on late-stage gate-level netlists, EFESOS enables early design space exploration aligned with ISO 26262 requirements. Experimental results on 125 graphs from diverse automotive-grade\\
architectures demonstrate that EFESOS achieves up to 93.2% accuracy and a minimum of 588× speedup compared to traditional FI campaigns. These results demonstrate its potential to accelerate FuSA analysis in complex SoCs using diverse GNN architectural variants, such as GCN, GAT and GraphSAGE.\\
\\
Ernesto Cristopher Villegas Castillo](/content/explore/84a77153-f80e-4784-b55d-dc0f817ec3d1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8299900/tree/v1/index.html)

[Biology\\
\\
3 \| Jun \| 2026\\
\\
Reproducible Computational Analysis for Advanced Dual Payload Antibody-Drug Conjugates\\
\\
This capsule contains the R scripts and reproducible computational environment (renv-locked) used for RNA-seq and pathway analysis in the study of Advanced dual payload antibody-drug conjugates for targeted cancer therapy.\\
\\
Zhuoxin “Zora” Zhou et al.](/content/explore/fce7e852-b8b2-41e6-b2fc-0ffcea8a93ac?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5962184/tree/v2/index.html)

[Computer Science\\
\\
23 \| Mar \| 2026\\
\\
An Attention-Enhanced Artificial Immune Classification Method with Dynamic Memory Evolution\\
\\
Inspired by the synergistic mechanisms of antigen presentation and homeostatic regulation in the biological immune system, A-AICM integrates a self-attention mechanism with an improved immune clonal selection theory to enhance feature extraction, classification performance, and generalization within high-dimensional complex data environments. Specifically, during the feature extraction stage, a self-attention mechanism is introduced to automatically identify and focus on key features from raw data, enhancing the representation of important information while suppressing redundant feature interference. In the immune response stage, a dynamic mutation operator adaptively adjusts the mutation magnitude based on stimulus intensity, enabling more efficient exploration of the optimal classification solution space. Finally, homeostatic regulation is achieved through an intelligent memory cell fusion mechanism, which compresses the memory pool size while maintaining high classification accuracy.\\
\\
Jiayi Zhang](/content/explore/58c43e45-0666-4436-ba65-7eef902b7bc8?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9862704/tree/v1/index.html)

[Bioinformatics\\
\\
10 \| May \| 2026\\
\\
Systematically decoding pathological morphologies and molecular profiles with unified multimodal embedding\\
\\
Multi-Embed is a unified and interpretable multimodal learning framework, which demonstrated superior performance in both cross-modality inference and integration of pathology morphologies and gene expression profiles. Multi-Embed enables the development of high-precision prognosis models, tissue architecture identification and the construction of spatiotemporal malignant trajectories across multiple cancer types, highlighting its great application potential.\\
\\
Chaofei Gao](/content/explore/f489f165-529e-4d0a-b864-0bf3ce433906?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4513986/tree/v2/index.html)

published in [Nature Methods](/content/explore?query=Nature%20Methods&refine=journal/index.html)

[Computer Science\\
\\
23 \| Mar \| 2026\\
\\
Attention-Augmented 3D U-Net for Murine Lung CT Segmentation\\
\\
This capsule provides the code and experimental configuration for the study \\
"Attention-Augmented 3D U-Net for Murine Lung CT Image Segmentation".\\
\\
The project implements a 3D U-Net architecture enhanced with attention gating \\
mechanisms and transfer learning initialization for segmenting murine lung CT \\
volumes. The dataset consists of 20 high-resolution murine CT scans provided by \\
the Royal College of Medicine, Cardiff.\\
\\
The capsule includes model training scripts, preprocessing pipelines, and \\
evaluation procedures used to generate the results reported in the paper.\\
\\
Key components include:\\
\\
- 3D U-Net with attention gates\\
- Transfer learning based encoder initialization\\
- Dice + weighted cross-entropy loss\\
- Patch-based training\\
- On-the-fly data augmentation\\
\\
The capsule is designed to reproduce the experiments described in the IEEE Access \\
submission and to support further research on data-constrained biomedical imaging tasks.\\
\\
Weiye Luo Cardiff University Department of Computer Science](/content/explore/cb9bb31a-4103-433b-8948-7d272d649c9d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4857692/tree/v1/index.html)

[Computer Science\\
\\
23 \| Mar \| 2026\\
\\
Carousel Greedy Julia implementation for the Minimum Vertex Cover Problem\\
\\
This capsule provides a reproducible Julia implementation of the Carousel Greedy algorithm applied to the Minimum Vertex Cover problem.\\
\\
The capsule contains the source code of the algorithm, a sample instance, and a script that executes the solver and reproduces the computational results.\\
\\
Running the capsule will automatically execute the algorithm on the provided graph instance and print the solution statistics.\\
\\
Contents:\\
\\
- Julia implementation of the Carousel Greedy algorithm\\
- Example graph instance\\
- Script to reproduce the experiment\\
\\
This capsule accompanies the article submitted to Software Impacts.\\
\\
Raffaele Dragone](/content/explore/820f7a08-d7f7-449b-886a-719186391cb7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2431091/tree/v1/index.html)

[Biology\\
\\
23 \| Mar \| 2026\\
\\
Butterfly Identification with Neural Networks\\
\\
This repository contains scripts for training neural networks for image classification – here images of Austrian butterflies collected by citizen scientists.\\
The scripts focus on distributed (HPC) systems, making use of data parallelism to speed up training on large datasets.\\
(A Reproducible Run here performs only an inference example with a trained model.)\\
\\
Andreas Lindner & Friederike Barkmann](/content/explore/16be22a9-b9dd-4b35-9fb8-1c71454a1669?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7431475/tree/v1/index.html)

[Medical Sciences\\
\\
23 \| Mar \| 2026\\
\\
Manifold Topological Deep Learning for Biomedical Data\\
\\
A Manifold Topological Deep Learning model for biomedical data classification\\
\\
Xiang Liu](/content/explore/d5743807-5700-4ffb-9caa-3c5efb0d37e7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5844659/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Computer Science\\
\\
20 \| Mar \| 2026\\
\\
Beyond Solidity: LLM-Assisted Vulnerability Detection on Solana and Algorand Smart Contracts](/content/explore/e2d1ab8c-9a29-4d05-8d70-5aa0d83870f5?page=2&filter=all/index.html)

Anmol Vats

[Open Capsule](/content/capsule/2504715/tree/v1/index.html)

[Computer Science\\
\\
17 \| Apr \| 2026\\
\\
Supplement for: VF-AdvJPEG for Efficient CPU-Based JPEG-Aware Transfer Attacks\\
\\
**VF-AdvJPEG Code Ocean Capsule** \\
\\
This directory is the code-only upload source for the VF-AdvJPEG Code Ocean capsule.\\
\\
It reproduces the canonical CPU study used for the IEEE SPL submission:\\
\\
- offline VF calibration\\
- `baseline_strong` and `vf_advjpeg`\\
- source `resnet18`, target `mobilenet_v2`\\
- four JPEG suites: `static_q70`, `static_q80`, `static_q90`, `dynamic_uniform`\\
- three seeds: `42`, `43`, `44`\\
\\
This capsule intentionally excludes manuscript, LaTeX, PDF, arXiv, and submission-packaging files.\\
\\
**Directory layout**\\
\\
- `src/vf_advjpeg`: runtime package subset only\\
- `scripts/run_code_ocean_repro.py`: single entrypoint used by Code Ocean `run.sh`\\
- `configs/code_ocean_capsule.yaml`: Code Ocean-specific canonical config\\
- `assets/pet37_ei_cpu_splits.json`: tracked split manifest\\
- `assets/canonical_expected_metrics.json`: regression baseline for the reproducible run\\
- `assets/data_assets_manifest.json`: expected `/data` contents and checkpoint checksums\\
- `assets/environment_lock.json`: starter environment lock, overlay package lock, and runtime policy\\
- `CODE_OCEAN_UPLOAD_GUIDE.md`: step-by-step upload and rerun instructions\\
\\
**Expected starter environment**\\
\\
- `PyTorch (2.4.0, CUDA 12.4.0, Mambaforge24.5.0-0, Python3.12.4, Ubuntu22.04)`\\
- The reproducible run is still CPU-only and sets `CUDA_VISIBLE_DEVICES=""`.\\
\\
**Environment note** \\
\\
The environment reported in the manuscript reflects the original local setup used for the submitted paper results. The Code Ocean capsule uses a different reproducible environment built from the currently available Code Ocean starter environment plus the pinned overlay dependencies included in this capsule.\\
\\
This is intentional. The goal of the capsule is to provide a stable, one-click executable reproduction of the canonical CPU study, not to mirror the authors' original local workstation exactly. The capsule is therefore intended to reproduce the manuscript's primary conclusions and canonical headline metrics under a platform-supported reproducibility environment.\\
\\
**Expected `/data`**\\
\\
- `/data/oxford-iiit-pet/images`\\
- `/data/oxford-iiit-pet/annotations`\\
- `/data/checkpoints/resnet18_pet37.pt`\\
- `/data/checkpoints/mobilenet_v2_pet37.pt`\\
- `/data/perceptual/alexnet-owt-7be5be79.pth`\\
\\
All generated outputs are written to `/results`.\\
\\
Ying Xu](/content/explore/dc2b1068-622f-4b7b-8a6c-6546da7f8fc2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8226743/tree/v1/index.html)

[Computer Science\\
\\
23 \| Mar \| 2026\\
\\
Carousel Greedy Matlab implementation for the Minimum Vertex Cover Problem\\
\\
This capsule provides a reproducible Matlab implementation of the Carousel Greedy algorithm applied to the Minimum Vertex Cover problem.\\
\\
The capsule contains the source code of the algorithm, a sample instance, and a script that executes the solver and reproduces the computational results.\\
\\
Running the capsule will automatically execute the algorithm on the provided graph instance and print the solution statistics.\\
\\
Contents:\\
\\
- Matlab implementation of the Carousel Greedy algorithm\\
- Example graph instance\\
- Script to reproduce the experiment\\
\\
This capsule accompanies the article submitted to Software Impacts.\\
\\
Raffaele Dragone](/content/explore/997857da-6271-4245-b005-0df6f4b6db91?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2923057/tree/v2/index.html)

[Social Sciences\\
\\
20 \| Mar \| 2026\\
\\
Predictive Analytics for Navigation Data Using Sequence-Based Clustering and Absorbing Markov Chains\\
\\
We propose to use absorbing Markov chain (AMC) and sequence-based clustering (SBC) to predict absorbing behaviors of human navigation on the Web. Unlike the Markov models that focus on predicting the next state behavior, AMC allows us to make predictions on the expected remaining navigation steps and the final state when navigation ends. Similarly, instead of estimating transition probabilities of a traditional Markov chain and then predicting the absorption behaviors, the proposed method opts to directly estimate the elements of the fundamental matrix of an AMC. In order to improve prediction performance, we adopt SBC to cluster navigation patterns and then estimate fundamental matrices for each cluster. In predicting absorption behaviors, the proposed method classifies each navigation pattern into one of the clusters formed by SBC and then applies the corresponding Markov model. We use two real-world navigation data sets to demonstrate the effectiveness of the proposed predictive analytics method, with various classification evaluation metrics including accuracy, $F\_1$-score, and area under the ROC curve. The proposed method is compared against other existing machine learning approaches to further highlight the strengths of our approach.\\
\\
Sungjune Park, Hyejin Ku & Richard Le](/content/explore/15d19c05-3f62-4ef5-953e-8e580642c0a7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0522427/tree/v1/index.html)

[Engineering\\
\\
20 \| Mar \| 2026\\
\\
Code for Reconstructing Whole-Brain Spatiotemporal Dynamics Using EEG/MEG Source Imaging with Geometric Constraints\\
\\
The brain-wide neural dynamics are shaped by the geometry of the brain. However, the mechanisms through which anatomical geometry constrains the organization and propagation of brain activity remain largely unmodeled, particularly at fast timescales. Here, we introduce a source imaging framework that incorporates geometric constraints derived from individual anatomical structure, enabling biologically plausible reconstructions of brain activity from EEG/MEG signals. By representing neural sources as combinations of geometric basis functions (GBFs) derived from cortical eigenmodes, this method aligns the inverse solution with the brain’s intrinsic geometry. As a result, the reconstructed activity reveals wave-like spatiotemporal patterns that reflect the geometry-guided organization of neural dynamics. We validate the framework across fMRI-informed synthetic benchmarks, cognitive tasks, resting-state recordings, intracranial stimulation, and epilepsy data. Compared to conventional methods, GBFs yield higher localization accuracy and capture spatiotemporal dynamic patterns consistent with anatomical pathways. Taken together, these findings suggest that brain activity—both spontaneous and evoked—can be parsimoniously modeled as excitations of subject-specific geometric modes. By linking cortical geometry to fast neural processes, this approach offers a unified and generalizable model for reconstructing brain-wide dynamics in both scientific and clinical applications.\\
\\
Song Wang et al.](/content/explore/11dfdfb7-bff3-4c9a-b1e0-81974a7d2f76?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8472672/tree/v1/index.html)

published in [Nature Biomedical Engineering](/content/explore?query=Nature%20Biomedical%20Engineering&refine=journal/index.html)

[Social Sciences\\
\\
19 \| Mar \| 2026\\
\\
Organizational mechanisms enable collective scientific cognition in AI agent teams\\
\\
Scientific knowledge creation is inherently a collective cognitive process, requiring the coordination of distributed expertise, structured interaction, and iterative refinement of ideas. Whether such collective scientific cognition can emerge within artificial systems composed of multiple AI agents, however, remains unclear. Here we introduce SciTown, a multi-agent scientist framework designed to embed organizational mechanisms within teams of large language model–based agents. Specifically, SciTown incorporates role differentiation to distribute domain expertise, structured multi-round debate to externalize and integrate disciplinary knowledge, and institutionalized peer review to introduce constructive conflict and refine evaluation criteria. Together, these mechanisms operationalize principles from knowledge creation theory and establish a coordinated organizational structure that guides hypothesis generation and verification planning. Through repeated cycles of interaction, tacit disciplinary intuitions are progressively transformed into explicit and integrated scientific strategies, enabling team-level reasoning to emerge from agent-level behaviors. Using autonomous grant proposal generation as a high-cognitive-load testbed, we evaluate the causal role of these organizational mechanisms through a three-tier experimental design, including SECI-based intervention experiments (N = 300), systematic ablation studies (N = 360), and human expert evaluations (N = 120). Across conditions, agent teams operating under structured organizational mechanisms produce proposals judged to be significantly more novel, coherent, and feasible than those generated by single-agent or unstructured multi-agent baselines. Together, these findings provide empirical evidence that collective scientific cognition can emerge in artificial agent teams when appropriate organizational mechanisms are embedded, offering new insights into the computational foundations of scientific collaboration and scalable approaches to augmenting human scientific productivity.\\
\\
Zhe Wang](/content/explore/dfd311b8-22cf-443d-8381-a3b5d26912e1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4190718/tree/v1/index.html)

[Computer Science\\
\\
19 \| Mar \| 2026\\
\\
Topology or Demography? Understanding Sources of Bias in Opinion Dynamics Models\\
\\
Ways in which people's opinions change are, without a doubt, subject to a rich tapestry of differing influences. Factors that affect how one arrives at an opinion reflect how they have been shaped by their environment throughout their lives, education, material status, what belief systems they subscribe to, and what socio-economic minorities they are a part of. This already complex system is further expanded by the ever-changing nature of one's social network. It is therefore no surprise that many models have a tendency to perform best for the majority of the population and discriminate against those people who are members of various marginalized groups. This bias and the study of how to counter it are subject to a rapidly developing field of Fairness in Social Network Analysis (SNA). \\
The focus of this work is to look into how a state-of-the-art model discriminates certain minority groups and whether it is possible to reliably predict for whom it will perform worse. Moreover, is such a prediction possible based solely on one's demographic or topological features? To this end, the NetSense dataset, together with a state-of-the-art CoDiNG model for opinion prediction, has been employed. Our work explores how three classifier models (Demography-Based, Topology-Based, and Hybrid) perform when assessing for whom this algorithm will provide inaccurate predictions. Finally, through a comprehensive analysis of these experimental results, we identify four key patterns of algorithmic bias: \\
\\
1. variability in predictive effectiveness across issue types and minority groups, \\
2. extreme predictive variability for behaviourally-defined minorities, \\
3. potent but variable predictive signals from socio-economic and religious minorities, \\
4. persistent challenges in accurately predicting misclassifications for certain demographic minorities. \\
Our findings suggest that no single paradigm provides the best results and that there is a real need for context-aware strategies in fairness-oriented social network analysis. We conclude that a multi-faceted approach, incorporating both individual attributes and network structures, is essential for reducing algorithmic bias and promoting inclusive decision-making.\\
\\
Stanisław Stępień & Michalina Janik](/content/explore/c1fa2791-b5a0-4987-8b3a-735ee7fbafbc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0333905/tree/v1/index.html)

[Physics\\
\\
19 \| Mar \| 2026\\
\\
Differential Cellulose Distribution Drives Polarized Growth of Cotton Fibers\\
\\
DiffCMFCottonFibers\\
\\
Wang Guangda et al. NCOMMS, 2025. Differential Cellulose Distribution Drives Polarized Growth of Cotton Fibers.\\
\\
Lvwen Zhou](/content/explore/243705d3-276a-4340-b587-d8c05e3018cf?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8398194/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
18 \| Mar \| 2026\\
\\
Anchor-Free Cooperative RSSI Localization for Disaster Rescue Using Sequential Monte Carlo and Cramér–Rao Analysis\\
\\
Complete MATLAB simulation code for the CoRescue-PF system — a cooperative particle-filter-based localization framework for locating disaster survivors using RSSI measurements from mobile rescuers, without requiring pre-deployed anchor infrastructure. Reproduces all figures and numerical results from the accompanying IEEE Sensors Journal paper: "Anchor-Free Cooperative RSSI Localization for Disaster Rescue Using Sequential Monte Carlo and Cramér–Rao Analysis."\\
\\
Md. Raihanul Islam Rahul](/content/explore/2f831055-d9d4-440e-9e5f-5a17d4c1f117?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1431056/tree/v1/index.html)

[Engineering\\
\\
18 \| Mar \| 2026\\
\\
Version \[2.0\]- VortexFitting: A post-processing fluid mechanics tool for vortex identification\\
\\
VortexFitting is a post-processing tool for identifying and characterizing vortices in fluid dynamics datasets. Version 2.0 introduces significant enhancements including a wxPython-based graphical user interface for improved usability, parallelization capabilities achieving up to 83% reduction in computation time, and expanded theoretical models (Rankine and Batchelor vortex models). High-order numerical schemes implemented via scipy improve accuracy and extensibility. The software has been validated on large-scale datasets from publicly available turbulence databases. These improvements make VortexFitting more accessible and efficient for analyzing complex flow structures in both experimental and computational fluid dynamics.\\
\\
Guilherme Anrain Lindner & Yann Devaux](/content/explore/d7a30d93-5c1c-4518-884c-3c1ec5aa33e8?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5450155/tree/v1/index.html)

[Physics\\
\\
18 \| Mar \| 2026\\
\\
Symmetry-guided catalogue of chiral phonon materials\\
\\
ChiralPY is a Python script designed to compute the chirality properties of phonons in crystalline materials.\\
\\
Yuanfeng Xu, Yue Yang, Yu Mao & Zhanghuan Li](/content/explore/d666d11f-67f1-496e-9831-f5f0747b099d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9518530/tree/v1/index.html)

published in [Nature Physics](/content/explore?query=Nature%20Physics&refine=journal/index.html)

[Biology\\
\\
17 \| Mar \| 2026\\
\\
CSF1R inhibition plus chemotherapy relieves systemic immune suppression in human metastatic triple negative breast cancer and sensitizes transgenic mammary tumors to anti-PD-1 therapy by relieving resident memory T cell dysfunction\\
\\
Data and code for analysis in manuscript "CSF1R inhibition plus chemotherapy relieves systemic immune suppression in human metastatic triple negative breast cancer and sensitizes transgenic mammary tumors to anti-PD-1 therapy by relieving resident memory T cell dysfunction"\\
\\
Abstract:\\
A significant correlation exists between presence of intratumoral macrophages, tumor progression, and poor outcomes in triple-negative breast cancer (TNBC) with limited therapeutic options available for advanced stage setting. Preclinical in vivo studies revealed that inhibition of myelomonocytic colony stimulating factor 1 (CSF1) or its receptor (CSF1R), plus cytotoxic therapy decreased primary tumor growth kinetics and pulmonary metastases by CD8+ T cell-dependent mechanisms. To translate these findings, we conducting a nonrandomized, open label phase 1b/2 study (NCT01596751) evaluating pexidartinib (PLX3397; PLX), a small molecule CSF1R inhibitor, in combination with eribulin mesylate, a non-taxane microtubule dynamics inhibitor, in patients with heavily pretreated metastatic TNBC. 12-week PFS was 36% (95% confidence interval (CI) 22.2% to 58.4%) with 44.8% of patients achieving clinical benefit with a small subset demonstrating prolonged disease control beyond 6 months. Correlative analyses in circulating blood from baseline revealed increased leukocyte activation including presence of CD8+ and CD4+ T memory cells, and PD-1 expression on CD4+ T cells in patients experiencing a partial response or stable disease. Leveraging clinical biomarkers of response, preclinical transgenic mammary carcinoma modeling identified synergy with PD-1 blockade, leading to a transient primary tumor regression in ~60% of the cohort, associated with expansion of effector and resident memory T cells. These clinical and preclinical findings together provide rationale for combinations to increase the therapeutic index for aPD-1 therapy by diminishing presence of T cell-suppressive myelomonocytic cells to further improve outcomes for patients with refractory disease.\\
\\
Amanda Poissonnier et al.](/content/explore/667eab52-63d2-46a7-8773-6f77f2fde5d7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5361499/tree/v1/index.html)

[Physics\\
\\
20 \| Mar \| 2026\\
\\
Interventional Process-Transport Cost: An Energy-Based Metric for Distinguishability Under Causal Feedback\\
\\
This capsule provides a fully reproducible framework for the simulation and estimation of the Interventional Process-Transport Cost (IPTC). The code implements a Stochastic Master Equation (SME) integrator for a superconducting transmon qubit under continuous homodyne monitoring. It evaluates the energetic cost of causal erasure, the control energy required to render the path law of an excited-state trajectory indistinguishable from the ground state, using a linearized Kalman filter as a causal estimator.The repository includes:A core simulation engine (iptc\_tsweep.py) performing a 240-configuration parameter sweep across coupling strengths ($\\chi$), measurement efficiencies ($\\eta$), and integration times ($T$).A visualization suite (plotting\_notebook.ipynb) that processes the raw simulated data to reconstruct the "SNR-degenerate" regimes and Pareto dominance transitions discussed in the associated manuscript.High-resolution vector outputs (PDF) of the manuscript figures. This work introduces IPTC as a resource-sensitive metric for quantum hardware optimization, complementing traditional passive distinguishability measures like SNR.\\
\\
Sinan Bugu](/content/explore/50176fce-de69-40b0-8b74-e2f2b3ff213d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0062864/tree/v1/index.html)

[Bioinformatics\\
\\
17 \| Mar \| 2026\\
\\
A minimal transcriptomic signature predicts intravascular tumor extension in renal cell carcinoma\\
\\
Renal cell carcinoma (RCC) with venous tumor thrombus, termed renal intravascular tumor extension (RITE), is associated with aggressive behavior and poor clinical outcomes. Yet, its underlying molecular determinants remain incompletely defined. We analyzed RNA sequencing data from three independent RCC cohorts comprising 721 samples. Two cohorts included matched samples of index tumor, tumor thrombus, and normal adjacent kidney tissue. Analyses integrated dimensionality reduction, differential gene expression, interpretable machine learning, and gene ontology approaches. Principal component analysis revealed that only these two cohorts exhibited a coherent RITE-associated transcriptional structure. Their sequencing depth was sufficient to delineate 6,317 differentially expressed genes that distinguish RITE from non-RITE tumors. SHAP-based feature attribution across logistic regression, random forest, and XGBoost yielded a robust 29-gene consensus signature, which was further distilled into a compact 13-gene panel that preserved maximal classification performance. These genes converged on biological themes, including loss of distal epithelial identity, dysregulation of ion transport pathways, and consistent enrichment of mitochondrial processes such as oxidative phosphorylation. Together, these findings define a newly discovered and uniquely refined molecular signature of venous tumor extension in RCC and highlight mechanistically relevant pathways that may inform biomarker development and future translational strategies for predicting or mitigating RITE progression.\\
\\
Christopher A. Mao et al.](/content/explore/7fe4a96d-cde5-48e8-988e-d44f891f2274?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8705245/tree/v1/index.html)

[Engineering\\
\\
17 \| Mar \| 2026\\
\\
Neuro-Optical Wolf Algorithm: A Feedback-Regulated Swarm Optimizer for Numerical and Engineering Optimization\\
\\
This paper proposes the Neuro-Optical Wolf Algorithm (NOWA), a swarm-based metaheuristic for numerical and engineering optimization that moves beyond single-metaphor design by introducing a feedback-regulated controller-operator architecture. Rather than relying solely on behavioral imitation, NOWA couples measurable search-state feedback with coordinated search operators to adapt exploration and exploitation online. Specifically, Adaptive Pupillary Response (APR) modulates search intensity using population diversity and recent fitness-improvement signals; Collective Motion Prediction (CMP) provides elite-centroid trajectory-aware guidance to accelerate convergence; Dichromatic Subspace Search (DSS) performs periodic sensitivity-guided dimensional focusing to reduce ineffective perturbations in higher-dimensional spaces; and Tapetum Lucidum Reflection (TLR) introduces a stagnation-triggered reflection operator that re-injects structured diversity while preserving elite guidance. A complexity analysis shows that the dominant internal computational overhead arises from APR’s diversity estimation, thereby clarifying the scalability trade-off with population size. NOWA is evaluated on the CEC-2022 single-objective bound-constrained benchmark suite at D=2, D=10, and D=20 over 30 independent runs, and on eight heterogeneous constrained engineering design problems. On CEC-2022, NOWA is most decisive at standard and higher dimensions, achieving the best overall Friedman mean rank at D=10 and D=20 (1.6667 in both cases), whereas the D=2 setting yields non-significant differences. On the engineering benchmark portfolio, NOWA attains the best overall mean rank (3.06) and records 35 wins out of 56 Holm-adjusted pairwise Wilcoxon comparisons (62.5%) against recent competitors. These results support the claim that the integrated NOWA architecture—combining feedback-regulated control, directional prediction, subspace focusing, and structured diversification—can improve robustness and competitiveness on rugged, high-dimensional optimization landscapes.\\
\\
ALLOUANI FOUAD et al.](/content/explore/f566f5bc-3932-442f-b6b2-ca68b1e0faa7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0880987/tree/v1/index.html)

[Engineering\\
\\
30 \| Mar \| 2026\\
\\
An anti-windup technique to optimize closed-loop post-saturation transients\\
\\
We present an anti-windup technique, independent of the controller structure, targeted at optimizing the transients that the controlled variable and the control input undergo when the latter exits saturation, which is an aspect where the available techniques offer hardly any control. \\
We formulate the problem as a constrained quadratic optimization one, and obtain for it an analytic solution.\\
This solution provides a value to which we reset the controller state on exiting saturation.\\
The resulting implementation is computationally lightweight, hence readily applicable to industrial controllers.\\
The reported simulation and experimental results evidence its ease of use, together with the yielded advantages.\\
\\
Andrea Bisoffi, Gian Paolo Incremona & Alberto Leva](/content/explore/d4bc9c40-af75-4afd-8f25-8e9593ed3b78?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6108228/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tcst.2026.3666689) published in [IEEE Transactions on Control Systems Technology](/content/explore?query=IEEE%20Transactions%20on%20Control%20Systems%20Technology&refine=journal/index.html)

[Engineering\\
\\
17 \| Mar \| 2026\\
\\
GFDFlow: an Object-Oriented Python Package for Modeling Two-Dimensional Transport Phenomena\\
\\
GFDFlow is an object-oriented Python package implementing the Generalized Finite Differences Method (GFDM) for 2D transport phenomena problems. Its key advantage is flexibility for handling irregular geometries without structured meshes. It supports Dirichlet and Neumann boundary conditions and targets applications in heat and mass transfer and fluid dynamics. The package is computationally efficient and accessible for both research and educational purposes.\\
\\
Ricardo Román-Gutiérrez, Carlos Chávez-Negrete & Francisco Domínguez-Mota](/content/explore/df163ea9-98a9-4eee-9dbe-26d974b7aa7e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7179462/tree/v1/index.html)

[Social Sciences\\
\\
2 \| Apr \| 2026\\
\\
Improving Small-Area Estimates of Public Opinion by Calibrating to Known Population Quantities\\
\\
Replication files for "Improving Small-Area Estimates of Public Opinion by Calibrating to Known Population Quantities" by William Marble and Josh Clinton ( _Political Analysis_).\\
\\
William Marble](/content/explore/530708ca-b9fe-4d9e-9277-72d426c753f5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9465544/tree/v2/index.html)

[Computer Science\\
\\
23 \| Mar \| 2026\\
\\
Carousel Greedy R implementation for the Minimum Vertex Cover Problem\\
\\
This capsule provides a reproducible R implementation of the Carousel Greedy algorithm applied to the Minimum Vertex Cover problem.\\
\\
The capsule contains the source code of the algorithm, a sample instance, and a script that executes the solver and reproduces the computational results.\\
\\
Running the capsule will automatically execute the algorithm on the provided graph instance and print the solution statistics.\\
\\
Contents:\\
\\
- R implementation of the Carousel Greedy algorithm\\
- Example graph instance\\
- Script to reproduce the experiment\\
\\
This capsule accompanies the article submitted to Software Impacts.\\
\\
Raffaele Dragone](/content/explore/94f8c2b9-f7d6-400a-8e39-28e7ef3a8664?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0430953/tree/v3/index.html)

[Computer Science\\
\\
23 \| Mar \| 2026\\
\\
Carousel Greedy Python implementation for the Minimum Vertex Cover Problem\\
\\
This capsule provides a reproducible Python implementation of the Carousel Greedy algorithm applied to the Minimum Vertex Cover problem.\\
\\
The capsule contains the source code of the algorithm, a sample instance, and a script that executes the solver and reproduces the computational results.\\
\\
Running the capsule will automatically execute the algorithm on the provided graph instance and print the solution statistics.\\
\\
Contents:\\
\\
- Python implementation of the Carousel Greedy algorithm\\
- Example graph instance\\
- Script to reproduce the experiment\\
\\
This capsule accompanies the article submitted to Software Impacts.\\
\\
Raffaele Dragone](/content/explore/b6df01bd-675e-45a1-aa12-02248bbf8998?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2190893/tree/v3/index.html)

[Engineering\\
\\
18 \| Mar \| 2026\\
\\
In-sensor wireless computing for intelligent remote sensing\\
\\
An in-sensor wireless computing architecture integrates imaging, compression and wireless transmission into one single step, dramatically reducing latency for large-field and high-resolution remote sensing applications.\\
\\
Yong Wang & Yuekun Yang](/content/explore/ec312dc1-d627-4e2f-837e-4da430900f64?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8892839/tree/v1/index.html)

[Associated article](https://doi.org/10.1038/s44460-026-00043-1) published in [Nature Sensors](/content/explore?query=Nature%20Sensors&refine=journal/index.html)

[Medical Sciences\\
\\
12 \| Mar \| 2026\\
\\
Patterns and Risks of Comorbidities among Children and Adults with Down Syndrome in the United States\\
\\
IMPORTANCE: There is a scant research exploring comorbidity patterns and altered risks of medical conditions specific to children and adults with Down syndrome (DS) in the United States (US). Using the largest US cohort of individuals with DS, this study aims to fill this research gap, informing precision healthcare and clinical practice.\\
OBJECTIVE: To identify comorbidity patterns among children and adults with DS and to quantify associations between DS and a full spectrum of medical conditions using both national insurance claims and regional Electronic Medical Record (EMR) data.\\
DESIGN, SETTING, AND PARTICIPANTS: This observational study used data from the Merative MarketScan database (2003-2023), containing de-identified insurance claims from over 200 million unique US residents, and an EMR dataset from a large Midwestern healthcare system. Individuals with DS were identified using the International Classification of Diseases, Ninth (ICD-9) diagnosis code of 758.0 (“Down syndrome”) or equivalent Tenth revision (ICD-10) diagnosis codes of Q90 (“Down syndrome”), Q90.0 (“Trisomy 21, non-mosaicism (meiotic nondisjunction)”), Q90.1 (“Trisomy 21, mosaicism (mitotic nondisjunction)”), Q90.2 (“Trisomy 21, translocation”), and Q90.9 (“Down syndrome, unspecified”). Up to ten individuals without DS were matched to each case by birth year (± 1 year), sex, and FIPS county code. Participants were categorized as children (< 18 years) or adults (≥ 18 years). The MarketScan cohort comprised 39,042 children and 22,130 adults with DS; the EMR validation cohort included 2,745 children and 3,325 adults with DS. We then conducted comorbidity pattern identification inspired by topic modeling and association analysis that regressed recurrence counts of medical conditions against DS status with sex and multiple age divisions as covariates.\\
EXPOSURES: For each individual in the two cohorts, we used variables including Down syndrome status, age, sex, and clinical diagnoses.\\
MAIN OUTCOMES AND MEASURES: We defined a comorbidity pattern as a recurrent combination of comorbid medical conditions across 1,569 diagnostic categories. We obtained the recurrence count of each medical condition for each individual by counting the total number of diagnoses for that condition over the individual’s health record. Multivariable regression models estimated associations between DS status and recurrence counts of each condition, adjusting for age and sex.\\
RESULTS: In the MarketScan cohort, 23 comorbidity patterns were identified in children and 17 in adults, with nine overlapping domains (“musculoskeletal,” “thyroid,” “sleep,” “heart,” “eye,” “ear,” “lung,” “convulsions,” and “respiratory” groups). Pediatric-specific patterns comprised “neuropsychiatric,” “language,” “aphasia,” “physiological development,” “symbolic,” “autism,” “incoordination,” “delayed milestones,” “muscle,” “developmental,” “brain,” “leukemia,” “immune,” and “dysphagia” groups. Adult-specific patterns included “neuritis,” “rhinitis,” “diabetes,” “renal,” “foot,” “urinary,” “digestion,” and “hypertension” groups. Among 1,569 evaluated medical conditions, individuals with DS in the MarketScan cohort had significantly increased risks for 776 conditions and decreased risks for 264 conditions compared with matched individuals without DS. External validation using the Midwestern EMR dataset replicated 401 increased-risk and 103 decreased-risk associations.\\
CONCLUSIONS AND RELEVANCE: In this large US cohort, children and adults with DS exhibited distinctive comorbidity patterns and altered comorbidity risks compared with matched individuals without DS. These findings provide a framework for risk stratification, clinical surveillance, and the development of evidence-based, lifespan-specific healthcare guidelines for individuals with DS.\\
\\
Gengjie Jia et al.](/content/explore/cd3a7f01-0c7e-4cfd-ab3f-e3ba05c54293?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5989099/tree/v1/index.html)

[Physics\\
\\
12 \| Mar \| 2026\\
\\
Collective Behavior and Memory States in Flow Networks with Tunable Bistability\\
\\
Code and Notebooks to produce the figures for: "Collective Behavior and Memory States in Flow Networks with Tunable Bistability"\\
\\
Multistability-induced hysteresis has been widely studied in mechanical systems, but such behavior has proven more difficult to reproduce experimentally in flow networks. Natural flow networks like animal and plant vasculature exhibit complex nonlinear behavior to facilitate fluid transport, so multistable flows may inform their functionality. To probe such phenomena in an analogous model system, we utilize an electronic network of hysterons designed to have tunable negative differential resistivity. We demonstrate our system’s capability to generate complex global memory states in the form of voltage patterns. The tunable nonlinearity of each element’s current-voltage characteristic allows us to navigate this space of hysteretic behavior, and can be utilized to engineer more complex networks with exotic effects such as avalanches and multiperiodic orbits.\\
\\
Experiments were performed in analog electronics using the MCCULW package. Simulations were performed using SPICE.\\
\\
Lauren Altman et al.](/content/explore/4706f466-73f0-4ae8-922e-1af22ce65e8e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9115676/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Medical Sciences\\
\\
11 \| Mar \| 2026\\
\\
PI3K and MAPK Signaling Nodes Serve as Divergent Drivers of Phenotypic Plasticity in Cancer-Associated Fibroblasts in Colorectal Cancer\\
\\
Cancer-associated fibroblasts (CAFs) exhibit phenotypic heterogeneity with each functional state playing critical roles in tumor progression. Notably, subtypes like inflammatory CAF (iCAF), characterized by increased chemokine/cytokine secretion, and myofibroblast-like CAF (myCAF), characterized by enhanced extracellular matrix (ECM) deposition and increased actomyosin contractility, can undergo phenotypic switching in response to cues from the tumor microenvironment (TME) and/or therapeutic interventions. However, the signaling pathways associated with their diverse phenotypes remain poorly understood. Through the analysis of single-cell RNA sequencing analysis of human colorectal cancer (CRC) we identified that the PI3K/mTOR and MAPK/ERK signaling pathways, among other pathways, are linked to the formation of myCAF and iCAF subtypes, respectively. Unbiased pharmacological interference of 12 distinct signaling pathways using three-dimensional (3D) human CRC-derived CAF cultures, ex vivo patient-derived tumor fragments and mouse models further revealed the significance of PI3K/mTOR and MAPK/ERK signaling in CAF plasticity and functional behavior. PI3K/mTOR inhibition drives iCAF formation through compensatory FGF-2 release and FGFR1–JAK2–STAT3 activation, leading to chemokine/cytokine secretion that promotes tumor spheroid growth and neutrophil infiltration. Conversely, MEK inhibition induces a myCAF phenotype via interferon-dependent ROCK and JAK1 signaling, resulting in ECM production that enhances tumor colony formation. In summary, our findings reveal a functional significance of respectively PI3K/mTOR and MAPK/ERK signaling pathways in CAF plasticity and underscore how standard-of-care targeted therapies can directly influence CAF phenotypes in CRC.\\
\\
Zihan Xia et al.](/content/explore/99920bba-6486-4f41-a923-8dde04b208f0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9909145/tree/v1/index.html)

[Engineering\\
\\
11 \| Mar \| 2026\\
\\
repository\[24\] of the paper "Elliptic jerk motion profile: nondimensional frequency-domain and time-domain analysis of second-order linear systems", IEEE trans. Mechatronics\\
\\
data repository \[24\] of the paper "Elliptic jerk motion profile: nondimensional frequency-domain and time-domain analysis of second-order linear systems", L. Bruzzone, D. Stretti, M. Verotti, P. Fanghella\\
\\
Luca Bruzzone](/content/explore/fcf1afc4-6ed6-4acf-8bd4-a60bd70fdbfe?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1968380/tree/v1/index.html)

[Computer Science\\
\\
11 \| Mar \| 2026\\
\\
AW-TPANet: A Clinical-Prior-Driven Dual-Task Network with Adaptive Windows for Upper Limb Rehabilitation Assessment\\
\\
This capsule contains the source code for the paper "AW-TPANet: A Clinical-Prior-Driven Dual-Task Network with Adaptive Windows for Upper Limb Rehabilitation Assessment", submitted to the IEEE Journal of Biomedical and Health Informatics (JBHI).\\
AW-TPANet is a cascaded dual-task framework that integrates Anatomically Constrained Spatial Encoding with Temporal Point Attention (TPA) and Adaptive Window (AW) mechanisms for simultaneous action classification and continuous standardness scoring.\\
Important Note on Data & Reproducibility:\\
Due to strict patient privacy and ethical restrictions, the original clinical dataset (RehabU-15) cannot be publicly shared. To demonstrate the reproducibility and execution pipeline of our code, we have provided a simulated dummy dataset () containing randomly generated tensors.fixed\_30frame\_dataset.pt\\
\\
The purpose of this Code Ocean capsule is to prove that the proposed AW-TPANet architecture is correctly implemented, and that the complete pipeline (data partitioning, model training) runs successfully without errors. Please note that the accuracy and loss metrics outputted in this run are based on the dummy data and do not reflect the actual state-of-the-art performance reported in our manuscript.\\
\\
DaWei Jiang](/content/explore/6de36e1d-fcb2-497f-b383-0f252f8987db?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4540178/tree/v1/index.html)

[Bioinformatics\\
\\
11 \| Mar \| 2026\\
\\
Early life exposure to N-nitrosamine drives genotoxicity, mutagenesis, and tumorigenesis in DNA repair-deficient mice\\
\\
**Data and Code Repository for Volk et al., 2025** \\
\\
This repository contains the source code for **Volk et al., 2025, Nature Communications**.\\
\\
**Project Overview** \\
\\
This study investigates the age-dependent susceptibility to liver carcinogenesis induced by N-Nitrosodimethylamine (NDMA). The project integrates multiple genomic and imaging approaches to understand why juvenile mice are significantly more susceptible to NDMA-induced tumors compared to adults.\\
\\
**Key Data Types** \\
\\
The manuscript includes data from four primary experimental approaches:\\
\\
1. **\[Genomics (Duplex Sequencing)\]**\\
\\
   - **Goal**: To measure mutagenesis with ultra-high accuracy.\\
   - **Data**: Mutation frequencies and mutational spectra analysis comparing NDMA-treated and control mice across age groups.\\
2. **Transcriptomics (RNAseq)**\\
\\
   - **Goal**: To identify differentially expressed genes and dysregulated pathways.\\
   - **Data**: Gene expression profiles focusing on inflammation, DNA repair, and interferon response pathways.\\
3. **Recombination Analysis (RaDR Mice)**\\
\\
   - **Goal**: To visualize and quantify homologous recombination events in situ.\\
   - **Data**: Fluorescence microscopy images of liver tissue from RaDR (Recombination and DNA Repair) reporter mice, quantifying both punctate foci (DNA damage response) and clonal expansions (mutagenic events).\\
4. **Tissue Pathology & Imaging**\\
\\
   - **Goal**: To assess tissue architecture, DNA damage, and cell proliferation.\\
   - **Data**:\\
     - **Histopathology**: H&E stained tissue sections for tumor burden and tissue morphology.\\
     - **Immunofluorescence**: Staining for DNA damage markers (pH2AX), cell proliferation (Ki67), and DNA replication (BrdU).\\
\\
Lindsay B Volk et al.](/content/explore/5434e211-66a1-4b7f-815e-80e602c457ab?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5824642/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Social Sciences\\
\\
11 \| Mar \| 2026\\
\\
Has transparency improved in clinical psychology? A repeated cross-sectional study (2012, 2018, 2024)\\
\\
Transparency is an important scientific principle that is frequently neglected in practice. The credibility crisis in psychology has catalysed new infrastructure, policies, and community initiatives focused on increasing the adoption of transparent research practices. However, it is unclear to what extent transparency has improved in clinical psychology. We conducted a repeated cross-sectional study to estimate the prevalence of transparent research practices in random samples of clinical psychology articles published in 2012, 2018, and 2024 (N = 589 articles). Funding and conflicts of interest disclosure statements were relatively common. By contrast, preregistration, sharing of measurement instruments, data, and analysis scripts, and use of reporting guidelines only increased modestly and remained uncommon overall. Authors provided few justifications for the lack of transparency. Overall, transparency in clinical psychology has increased, but considerable scope remains for improvement. Continued action is needed to realise a scientific ecosystem where transparency is the norm rather than the exception.\\
\\
Tom E. Hardwicke](/content/explore/5aae4013-2d85-4608-89bb-9332bc810f36?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2823062/tree/v2/index.html)

[Medical Sciences\\
\\
9 \| Mar \| 2026\\
\\
RACING: An open R implementation of validated linear and nonlinear gait analysis algorithms for triaxial accelerometer data\\
\\
RACING (R-based Attractor Complexity INdex for Gait) is an open R implementation\\
of validated nonlinear gait analysis algorithms originally developed in MATLAB for\\
the ACIER study. This capsule processes triaxial lower-back accelerometer recordings\\
(CSV format, 256 Hz) to extract gait stability and complexity metrics: Local Divergence\\
Score (LDS), Attractor Complexity Index (ACI), step/stride regularity, and step\\
frequency. The pipeline demonstrates end-to-end processing of the ACIER outdoor\\
older adult dataset (available on Zenodo, doi:10.5281/zenodo.10148824), producing\\
a fully annotated Excel results file. Developed at HE-Arc Sante Neuchatel under\\
the HES-SO ORD 2025 grant following FAIR principles.\\
\\
Philippe Terrier](/content/explore/a405f445-36ed-47f9-b94f-165c5ab2e61e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5177613/tree/v1/index.html)

[Earth Sciences\\
\\
9 \| Mar \| 2026\\
\\
Photolytic oxidation of ammonium chloride as a source of Cl2 in the atmosphere](/content/explore/24b699db-bd22-4c64-a800-407bb34d015b?page=2&filter=all/index.html)

[This model is largely based on the open-source Framework for 0-D Atmospheric Modeling (F0AM v4.3.0.1,](/content/explore/24b699db-bd22-4c64-a800-407bb34d015b?page=2&filter=all/index.html) [https://github.com/wolfegm/F0AM](https://github.com/wolfegm/F0AM)
).The complete model framework and core routines are available from the F0AM GitHub repository, with copyright information stored in the Docs section. The original model structure and configuration are retained. Modifications in this work are limited to the chemical mechanism, where additional chlorine-related reactions were implemented, and the model is run using the standard F0AM workflow.
The simulation shown here represents one example case used to evaluate the impact of the proposed mechanism on Cl2 production. For the assessment of Cl2 production from known mechanisms under field conditions, a similar modeling workflow and contribution analysis procedure are applied.

Shuying Li

[Open Capsule](/content/capsule/4100754/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Physics\\
\\
7 \| Mar \| 2026\\
\\
Unsupervised manifold learning of topological order derived from quantum circuit complexity](/content/explore/dd3a1653-2c08-4fdf-8b96-3dde57c63c90?page=2&filter=all/index.html)

[The codes provide unsupervised clustering of topological quantum phases / orders of bond alternating XXZ spin chain, Kitaev's toric code ground state, etc., derived theoretically from Nielsen's quantum circuit complexity; Data of ground states are calculated from the MPS-based DMRG using the TeNPy Library (](/content/explore/dd3a1653-2c08-4fdf-8b96-3dde57c63c90?page=2&filter=all/index.html) [https://github.com/tenpy/tenpy](https://github.com/tenpy/tenpy)).
Note: Read README in each code file before running the codes.

Yanming Che, Clemens Gneiting, Xiaoguang Wang & Franco Nori

[Open Capsule](/content/capsule/0455844/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
6 \| Mar \| 2026\\
\\
Mesoscale Carbon Fiber Lattices with Foam-Like Weight and Bulk Strength\\
\\
Supplementary Code 1. Python script implementing a genetic algorithm to compute shortest continuous fiber paths in SC and FCC lattices from 3D node coordinates, used for fabrication path planning.\\
\\
Supplementary Code 2. Optimization code for lattice beam stiffness using experimental SC, DC, and FCC bending slopes, identifying lightweight (<8 g) configurations with target performance.\\
\\
Jun Young Choi](/content/explore/4ded913a-3113-4ade-b2dc-507d78bc6496?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0533164/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Bioinformatics\\
\\
6 \| Mar \| 2026\\
\\
Mutation Reporter: Protein-Level Identification of Single and Compound Mutations in NGS Data](/content/explore/a3e24786-0a55-4f38-b30d-5773a55a00a0?page=2&filter=all/index.html)

[Next-generation sequencing (NGS) has accelerated precision medicine by enabling simultaneous analysis of multiple genes and detection of low-frequency mutations. However, few open-source tools allow non-specialized users to transparently adjust quality parameters during mutation analysis. Mutation Reporter was developed to identify both single and compound amino acid alterations directly from raw fastq files of sequencing originated from RNA or exon sequences. The software provides full parameter control—including alignment e-value, minimum read length, minimum read depth, and minimum variant allele frequency (VAF). The software is freely available under a GNU license on GitHub (\\texttt{](/content/explore/a3e24786-0a55-4f38-b30d-5773a55a00a0?page=2&filter=all/index.html) [https://github.com/meidanis-lab/mutation-reporter}](https://github.com/meidanis-lab/mutation-reporter%7D)).

Mikaela Teodoro et al.

[Open Capsule](/content/capsule/3930393/tree/v1/index.html)

[Earth Sciences\\
\\
6 \| Mar \| 2026\\
\\
Disequilibrium response to tapping crustal magma reveals storage conditions\\
\\
This code expands previous work by Coumans et al. (2020) to solve for diffusion of water vapor and CO2 into vapor bubbles, along with reaction between molecular water and hydroxyl groups in silicate melts.\\
\\
Janine Birnbaum et al.](/content/explore/79c5948e-d09d-43eb-a202-25c0dc3dc1cc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9205921/tree/v1/index.html)

published in [Nature](/content/explore?query=Nature&refine=journal/index.html)

[Earth Sciences\\
\\
6 \| Mar \| 2026\\
\\
A Snow-Fire Bridge Mechanism for 2025 Southern California Winter Wildfire\\
\\
This capsule provides the full codebase used in the manuscript, including both the core analysis routines and the plotting scripts used to generate the figures in the main text. It includes complete implementations of key dynamical diagnostics, such as: Rossby wave activity flux (WAF),\\
Eddy kinetic energy (EKE), Geopotential tendency equation diagnostics.\\
All computational components are organized into modular, reusable units to ensure transparency, reproducibility, and ease of application in related dynamical studies. The main dynamical‑diagnostic logic has been further refactored into modular, reusable functions to improve clarity, reproducibility, and ease of integration into related dynamical studies.\\
\\
Shizuo Liu](/content/explore/287d4357-8170-4030-99a1-7fa79380bc91?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5802944/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Computer Science\\
\\
5 \| Mar \| 2026\\
\\
Data-In-situ Computing with One-Pixel-Multiple-Memristor Architecture for Neuromorphic Sequential Vision\\
\\
A demo of data-in-situ computing network for MNIST classification\\
\\
Jiangrong Shen, Yi Sun & Wei Wang](/content/explore/e05f1ee0-25fd-4a42-a862-cb065b31049f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3537141/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Mathematics\\
\\
5 \| Mar \| 2026\\
\\
Finite-time convergence to an ϵ-efficient Nash equilibrium in potential games\\
\\
This repository accompanies the paper:\\
\\
Maddux, A., Ouhamma, R., Catic, H., and Kamgarpour, M., "Finite-time convergence to an ϵ-efficient Nash equilibrium in potential games", IEEE Transactions on Control of Network Systems (2026).\\
\\
This paper investigates the convergence time of log-linear learning to an ϵ-efficient Nash equilibrium in\\
potential games, where an efficient Nash equilibrium is defined as the maximizer of the potential function. Previous literature provides asymptotic convergence rates to efficient Nash equilibria, and existing finite-time rates are limited to potential games with further assumptions such as the interchangeability of players.We prove the first finitetime convergence to an ϵ-efficient Nash equilibrium in general potential games. Our bounds depend polynomially on 1/ϵ, an improvement over previous bounds for subclasses of potential games that are exponential in 1/ϵ. We then strengthen our convergence result in two directions: first, we show that a variant of log-linear learning requiring a\\
constant factor less feedback on the utility per round enjoys a similar convergence time; second, we demonstrate the robustness of our convergence guarantee if log-linear learning is subject to small perturbations such as alterations in the learning rule or noise-corrupted utilities.\\
\\
Anna Maddux, Reda Ouhamma, Hana Catic & Maryam Kamgarpour](/content/explore/8b2cc13b-3049-485e-be1c-88bd3a470f50?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2276148/tree/v1/index.html)

[Earth Sciences\\
\\
5 \| Mar \| 2026\\
\\
Desert dust exerts twice the longwave radiative heating estimated by climate models\\
\\
This code obtains the DustCOMM data set, including a calculation of the longwave direct radiative effect of desert dust. The main function to run DustCOMM is main\_constrain\_modern\_dust\_cycle.m. This function takes the following inputs: \\
\\
• The number of desired bootstrap iterations (e.g., ~50 for quick results; 1000 iterations are normally used for results used in a publication) \\
\\
• The model # to use (this should almost always be zero) \\
\\
• A spare variable that can be used for various purposes (this should almost always be zero) \\
\\
• The year for which to run the code. This should be 0 to run the code for the standard 2004-2008 climatology obtained in Kok et al. (2021a, b). \\
\\
The main\_constrain\_modern\_dust\_cycle.m code performs the following actions: \\
\\
- It initializes variables and settings. Some of this is done in the sub-scripts load\_data\_and\_pars\_modern\_dust\_cycle.m and initialize\_variables.m \\
\\
- It loops over the number of desired bootstrap iterations (the variable nBoots). For each bootstrap iteration, it randomly draws a globally averaged atmospheric dust size distribution (from Adebiyi and Kok, 2020), size-resolved mass extinction efficiency (based on Kok et al., 2017), and a set of global model simulations (see pick\_model.m) of how 1 Tg of dust loading of each particle size from each of the 9 dust source regions is distributed across the world (from the CESM, IMPACT, GISS ModelE2.1, GEOS/GOCART, MONARCH, or INCA models). \\
\\
- It then draws the dust AOD at the 15 regions defined in Ridley et al. (2016) (see obtain\_Ridley\_DAOD.m) \\
\\
- Then it figures out how many Tg of dust are needed from each of the 9 source regions to produce a 2D map of dust AOD that minimizes the disagreement against the dust AOD in the 15 regions. \\
\\
- Copy\\
\\
```\\
It then calculates the various attributes of the global dust cycle, including the LW direct radiative effect (in the function calc_LW_rad_forcing.m, which is called by the script calc_results_modern_dust_cycle.m)\\
```\\
\\
- The processing of the results of each bootstrap iteration occurs in process\_results\_modern\_dust\_cycle \\
\\
\\
The abstract of the corresponding article, titled " **Desert dust exerts twice the longwave radiative heating estimated by climate models**", is as follows: \\
\\
Although desert dust is the most abundant atmospheric aerosol by mass, its longwave radiative effects remain unclear, obscuring the impacts of dust on weather and climate. Using a data-driven analytical model constrained by observations, we show that scattering and absorption of longwave radiation by dust heats the planet by +0.25 ± 0.06 W m⁻² (90% confidence). This is nearly twice the value simulated by current climate models, which omit longwave scattering and underrepresent super coarse dust (diameter > 10 μm). These omissions bias modeled surface energy fluxes, cloud responses, precipitation, and atmospheric circulation. At the global scale, the sign and magnitude of the net dust direct radiative effect remain uncertain, with additional work needed to constrain shortwave cooling effects. These findings show that improving the representation of dust interactions with longwave radiation can improve weather forecasting and is essential to resolve the role of dust in climate change.\\
\\
Jasper F. Kok](/content/explore/34a601c8-cc98-439c-b779-18c353b9231c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6254038/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Bioinformatics\\
\\
5 \| Mar \| 2026\\
\\
ETSAM: Effectively Segmenting Cell Membranes in cryo-Electron Tomograms\\
\\
Cryogenic Electron Tomography (cryo-ET) is an emerging experimental technique to visualize cell structures and macromolecules in their native cellular environment. Accurate segmentation of cell structures in cryo-ET tomograms, such as cell membranes, is crucial to advance our understanding of cellular organization and function. However, several inherent limitations in cryo-ET tomograms, including the very low signal-to-noise ratio, missing wedge artifacts from limited tilt angles, and other noise artifacts, collectively hinder the reliable identification and delineation of these structures. In this study, we introduce ETSAM - a two-stage Segment Anything Model 2 (SAM2)-based fine-tuned AI method that effectively segments cell membranes in cryo-ET tomograms. It is trained on a diverse dataset comprising 83 experimental tomograms from the CryoET Data Portal (CDP) database and 28 simulated tomograms generated using PolNet. ETSAM achieves state-of-the-art performance on an independent test set comprising 10 experimental tomograms for which ground-truth annotations are available. It robustly segments cell membranes with high sensitivity and precision, significantly outperforming existing deep learning methods.\\
\\
Joel Selvaraj & Jianlin Cheng](/content/explore/0020e296-aa81-4f61-a0ed-c6df35e7df33?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5996623/tree/v1/index.html)

[Associated article](https://doi.org/10.1101/2025.11.23.689996) published in [bioRxiv](/content/explore?query=bioRxiv&refine=journal/index.html)

[Biology\\
\\
5 \| Mar \| 2026\\
\\
Resistance and resilience of biodiversity in a tropical rainforest\\
\\
This code calculates recoverytimes, resistance and resilience of biodiversity from data sampled in Reserva Canandé, Ecuador. Furthermore, all figures that are part of the corresponding papers are created with this code.\\
\\
Timo Metz et al.](/content/explore/56c79e12-efb3-4520-82ac-e60e787ce148?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2113682/tree/v1/index.html)

published in [Nature](/content/explore?query=Nature&refine=journal/index.html)

[Bioinformatics\\
\\
5 \| Mar \| 2026\\
\\
Supplementary code for: The Linkage between Microbial Community Dynamics and Urbanization Age.\\
\\
Supplementary code for: The Linkage between Microbial Community Dynamics and Urbanization Age.\\
\\
Yinghui Jia, Jun Wu, Lan Wang & Tieliu Shi](/content/explore/db557144-09b2-4e30-b65b-e77ebee2903c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9529451/tree/v1/index.html)

published in [Nature Sustainability](/content/explore?query=Nature%20Sustainability&refine=journal/index.html)

[Computer Science\\
\\
4 \| Mar \| 2026\\
\\
Real-Time Kill Chain State Machine (RT-KCSM)\\
\\
Cyber threats present an ongoing challenge for organisations worldwide. Attackers range from cybercriminals to state-funded groups that have a specialised skill set to execute complex attacks and present an Advanced Persistent Threat (APT). Therefore, organisations use security monitoring as a second line of defence to detect attacks based on signatures that raise alarms when an Indicator of Compromise (IoC) is observed. However, current Intrusion Detection Systems (IDS) generate many false positives, leading to alert fatigue. The raised alerts also do not show the whole attack as they need to be reviewed individually. Our work presents a simplified approach that enables efficient and real-time construction of attacks by correlating alerts. Tests with different network datasets suggest that our prioritisation mechanisms can reduce the number of false-positive alerts by 99 %. Our performance evaluation indicates that we can detect multi-stage attacks in real-time with a low memory footprint and short execution time.\\
\\
Liliana Kistenmacher, Anum Talpur & Mathias Fischer](/content/explore/4a29d0bd-41f8-458b-8fdf-b1fd60922a6c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2288451/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/cns66487.2025.11194951) published in [2025 IEEE Conference on Communications and Network Security (CNS)](/content/explore?query=2025%20IEEE%20Conference%20on%20Communications%20and%20Network%20Security%20(CNS/index.html)&refine=journal)

[Computer Science\\
\\
4 \| Mar \| 2026\\
\\
RIS, A Topology-Preserving Framework for Adaptive Transformer Sparsification\\
\\
**Code Ocean Capsule: Reduced Interaction Sampling (RIS) Framework**  **Overview** \\
\\
This compute capsule provides the reference implementation, high-performance benchmarking suite, and reproducible figure-generation scripts for the **Reduced Interaction Sampling (RIS)** framework.\\
\\
RIS is a stochastic sparsification algorithm designed to break the O(N2)O(N^2)O(N2) quadratic complexity bottleneck inherent in self-attention mechanisms and large-scale graph analytics. By exploiting the deep informational redundancy present in complex networks, the Heuristic Adaptive Function (HAF) within RIS identifies a "structural skeleton" of high-degree hubs and bridging edges. This allows algorithms to transition from O(N2)O(N^2)O(N2) to O(Nlog⁡N)O(N \\log N)O(NlogN), or even to linear O(N)O(N)O(N), operations while maintaining global topological fidelity.\\
\\
**Key Reproducible Results** \\
\\
This capsule automatically reproduces the core empirical findings of the manuscript:\\
\\
1. **The Quadratic Wall Break**: Achieving near-total reduction of redundant pairwise operations for large networks, validating the theoretical shift to linear complexity. \\
2. **Topological Preservation**: Generating Pearson correlations between true vs. sampled degree centrality on synthetic scale-free networks, maintaining ρ>0.75\\rho > 0.75ρ>0.75 through a stable Transition Phase.\\
3. **Extreme Sparsity Performance (Transformers)**: Simulating context-window patterns on the 4-million-node com-LiveJournal network at ≈0.06%\\approx 0.06\\%≈0.06% density (the scale of modern frontier LLMs), where RIS recovers 4.4% more global hubs than Longformer and BigBird across 20,000 statistical replicates.\\
4. **Universality Across Domains**: Validating RIS behavior across scientific collaboration arrays, financial trust graphs, and global air transport networks vs. regular power grid lattices.\\
\\
**Execution** \\
\\
The default `run` script automatically detects the environment and executes the Python 3 orchestration suite to process the synthetic and real-world network data. \\
\\
All final outputs—including the CSV metrics and high-resolution PDFs mirroring the manuscript's Figures 1 through 5—are generated natively and exported to the `/results` directory.\\
\\
Federal University of Uberlândia](/content/explore/f03b987c-d03a-4517-ae43-0060183618ec?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9674542/tree/v1/index.html)

[Computer Science\\
\\
4 \| Mar \| 2026\\
\\
CRec: A tool for recovering the compilability of OCRed code\\
\\
CRec is a tool for fixing lexical and syntactic errors in OCRed code. The goal of CRec is to recover the compilability of OCRed code, thereby making the output directly usable for developers.\\
\\
anonym1610](/content/explore/f17e5350-0f21-4527-ae98-657ed03e18e9?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6956282/tree/v1/index.html)

[Computer Science\\
\\
3 \| Mar \| 2026\\
\\
patch-hub: A terminal-based tool to streamline Linux patch reviews\\
\\
The capsule setups patch-hub source code. patch-hub is a terminal based software written in Rust that streamlines a core workflow of the Linux kernel development model: patch-review\\
\\
Lorenzo Bertin Salvador, David Tadokoro, Hannah Harrisonn & Paulo Meirelles](/content/explore/cd010fd2-32fb-4b5f-bf4c-a0a0c1b54948?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9126139/tree/v1/index.html)

[Biology\\
\\
2 \| Mar \| 2026\\
\\
Botero 2026. The evolutionary consequences of behavioral plasticity\\
\\
R script to reproduce all simulations used in "The evolutionary consequences of behavioural plasticity". The code will create new simulations and will produce new data figures included in the paper.\\
\\
Carlos A. Botero](/content/explore/e6f0afc9-8ed1-4b68-9140-98538b9aa12f?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9952373/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Computer Science\\
\\
2 \| Mar \| 2026\\
\\
Evaluation Code for the FastReChain Topology Engineering Algorithm of OCS-Based Clusters\\
\\
The evaluation code for the paper "FastReChain: A Novel Bidirectional Model-Based Algorithm for Topology Engineering of OCS-Based Clusters".\\
\\
Zihan Zhu](/content/explore/4982c922-7586-43d2-898d-80171c244144?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1850532/tree/v1/index.html)

[Computer Science\\
\\
2 \| Mar \| 2026\\
\\
Flexible and Efficient Feature-level Fusion with Wireless Acoustic Sensors using Graph Attention Networks](/content/explore/4a36f267-4124-4bdf-bbdb-0a921dd2202e?page=2&filter=all/index.html)

[**WASN GNN** \\
This repository contains the implementation of manuscript `Flexible and Efficient Feature-level Fusion with Wireless Acoustic Sensors using Graph Attention Networks`.\\
**Installation** \\
Install dependencies from `requirements.txt`:\\
\\
bash\\
\\
Copy\\
\\
```bash\\
pip install -r requirements.txt\\
```\\
\\
**Project Structure**\\
\\
- **Data Preprocessing**: Scripts located in `/data` directory\\
- **Training**: Scripts prefixed with `train_`\\
- **Evaluation**: Scripts prefixed with `eval_`\\
\\
**Quick Start**  **Download Data**](/content/explore/4a36f267-4124-4bdf-bbdb-0a921dd2202e?page=2&filter=all/index.html)

[Download the SINS domestic activity dataset (](/content/explore/4a36f267-4124-4bdf-bbdb-0a921dd2202e?page=2&filter=all/index.html) [https://github.com/KULeuvenADVISE/SINS\_database?tab=readme-ov-file#download](https://github.com/KULeuvenADVISE/SINS_database?tab=readme-ov-file#download)) and place it in the `/data` directory.

Then follow these steps:

1. Adjust data dir variable in preprocessing scripts from `/data` and run it, the wav version is needed for the GCC-PHAT baseline from Kawamura et al.
2. Train and save the pretrained classifier using `train_frozen_classifier.py`.
3. Execute training of different sensor fusion strategy with `train_swarm_*.py`, the test performance is shown at the end of the training and stored as experiment files, the GNN framework (ours) allows the evaluation using a variable number of sensors, while the others can only use the same sensor node combination as the one used during training.
4. Evaluate the heuristic based decision level baseline models that can accept a vairable number of sensors using `eval_decision_level_baselines.py`.
5. Evaluate the ablation results: inference time and parameters/Macs using the corresponding `eval_*.py`.

Wei Wei

[Open Capsule](/content/capsule/6659699/tree/v1/index.html)

[Engineering\\
\\
2 \| Mar \| 2026\\
\\
Optimal Transport-Based Decentralized Multi-Agent Distribution Matching\\
\\
Official implementation of the paper "Optimal Transport-Based Decentralized Multi-Agent Distribution Matching", published in IEEE Transactions on Automatic Control (TAC), 2026.\\
\\
Kooktae Lee](/content/explore/43841e75-4d71-4968-b53c-53e0b57471a1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1956120/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tac.2026.3668445) published in [IEEE Transactions on Automatic Control](/content/explore?query=IEEE%20Transactions%20on%20Automatic%20Control&refine=journal/index.html)

[Social Sciences\\
\\
2 \| Mar \| 2026\\
\\
Monotone Ecological Inference\\
\\
We study the identification of individual-level associations when only aggregate data is available. We characterize the biases of, and relationships among, canonical ecological inference estimators. We use these results to develop a partial identification approach: monotone ecological inference. The approach exploits information about one or both of the following conditional associations: (1) outcome differences between groups within the same neighborhood, and (2) outcome differences within the same group between neighborhoods with different group compositions. We show how assumptions about the sign of these conditional associations, whether individually or in relation to one another, can yield informative sharp bounds. We illustrate our results using county-level data to study differences in Covid-19 vaccination rates among Republicans and Democrats in the United States.\\
\\
Hadi Elzayn et al.](/content/explore/22c41f76-af7d-46c9-8307-e224209f0dab?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4440816/tree/v1/index.html)

[Bioinformatics\\
\\
2 \| Mar \| 2026\\
\\
Cross-Site Portability of Parkinson’s Disease Microbiome Classifiers Under Distribution Shift\\
\\
This repository contains the complete reproducible analysis pipeline supporting the manuscript:\\
\\
“Site-Level Distribution Shift Dominates Disease Signal and Undermines Cross-Site Generalization in High-Dimensional Microbiome Classifiers.”\\
\\
The project evaluates transportability of microbiome-based Parkinson’s disease (PD) classifiers across three geographically distinct 16S rRNA cohorts (Finland, Malaysia, USA; total n = 682). Genus-level abundance tables were harmonized into a shared feature space of 108 genera and analyzed under centered log-ratio (CLR) transformation.\\
\\
Ridge-penalized logistic regression models were evaluated using strict within-cohort cross-validation and leave-one-cohort-out (LOCO) validation to quantify performance degradation under cross-site distribution shift.\\
\\
Key results:\\
\\
High internal discrimination (AUC up to 0.944).\\
\\
Substantial degradation under LOCO validation (AUC as low as 0.506).\\
\\
Cohort identity explains R² = 0.689 of compositional variance.\\
\\
PD status explains R² = 0.0092.\\
\\
Disease-associated effect vectors are nearly orthogonal to the dominant ecological axis (PC1), explaining geometric instability of cross-site generalization.\\
\\
The repository includes harmonized genus-level matrices, metadata, full analysis scripts (C21–C33), performance tables, structural alignment metrics, and all manuscript figures. All results are fully reproducible using R (version 4.4.2) with deterministic random seeds.\\
\\
This archive enables exact computational replication of the reported findings and supports transparent evaluation of distributional robustness in microbiome-based biomedical AI models.\\
\\
Antonio Pereira, Jessica Gama & Bianca Neves](/content/explore/a0977fa9-8911-4087-8237-0196e67885d5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3547455/tree/v1/index.html)

[Biology\\
\\
2 \| Mar \| 2026\\
\\
Adaptive optical correction for in vivo two-photon fluorescence microscopy with neural fields\\
\\
This repository contains the code for Neural fields for Adaptive Optical Two-photon Fluorescence Microscopy (NeAT). NeAT operates in three stages:\\
\\
1. **Aberration and structural estimation**\\
\\
From a single 3D image stack, NeAT estimates system or sample aberrations and recovers the underlying structural information, without any external training data.\\
\\
2. **Conjugation error correction**\\
\\
NeAT estimates and corrects conjugation errors in the imaging system, typically caused by incomplete conjugation or alignment errors.\\
\\
3. **Sample motion correction**\\
\\
NeAT maintains performance even with sample motion during the acquisition of the 3D input stack, by adaptively registering and correcting slice-to-slice motion artifacts.\\
\\
\\
* * *\\
\\
**Code overview** \\
\\
This capsule includes three main scripts:\\
\\
**`neat_learning.py`** \\
\\
Core routine that takes a 3D image stack as input and outputs aberration and structural estimations. Key physical parameters can be set via command-line arguments (a full list of tunable options is in the `args` parser):\\
\\
- `psf_dx, psf_dy, psf_dz`: pixel sizes (µm) along xxx, yyy, zzz axes \\
- `cnts`: center coordinates of the input stack \\
- `dims`: dimensions of the input stack \\
- `na_exc`: excitation numerical aperture (NA) \\
- `sample_motion`: Boolean parameter for sample motion correction. True if sample motion correction is performed.\\
\\
**Expected outputs:**\\
\\
- **`rec.h5`**— HDF5 file containing:\\
  - `out_x_m`: estimated structure \\
  - `out_k_m`: estimated PSF \\
  - `out_y`: computed image stack \\
  - `wf`: Zernike coefficients \\
  - `loss_list`: losses per epoch \\
  - `y`: normalized input stack \\
  - `y_min`, `y_max`: normalization bounds\\
- **`est_aber_map.bmp`** — bitmap of the estimated aberration map from `wf` (unit: waves)\\
- **`slm_pattern.bmp`** — 8-bit grayscale image (with a pixel value of 255 corresponding to 1 wave) that is the wrapped corrective phase pattern to be applied to an SLM for aberration correction, if needed.\\
\\
Feel free to explore the code and adjust parameters to suit your imaging setup.\\
\\
**`neat_conj_est.py`** \\
\\
This script estimates conjugation errors that may be present in a two-photon imaging system. It should be run after `neat_learning.py` on six image stacks. For testing, example stacks acquired using a microscope at UC Berkeley are available under `/commercial_beads_zeros/`, `/commercial_beads_mode4/`, `/commercial_beads_mode6/`, … and `/commercial_beads_mode13/`. Users should acquire these image stacks using their own system to estimate conjugation errors specific to their microscope. To acquire these stacks, users need to display on the SLM images corresponding to a flat phase pattern and five different Zernike modes for calibration (modes 4, 6, ..., 13), respectively. These patterns are available in `/slm_patterns_for_calibration/` as `k_vis_zeros.bmp`, `k_vis_mode4.bmp`, `k_vis_mode6.bmp`, ..., `k_vis_mode13.bmp.` (To apply these patterns directly on a SLM, the SLM needs to be calibrated so that a pixel value of 255 corresponds to 1 wave. The patterns can also be scaled and formatted for a deformable mirror.) Conjugation errors are computed as an affine transformation H^\\hat{H}H^, from the reconstructions of these stacks and saved to `H.h5`.\\
\\
**`neat_conj_corr.py`** \\
\\
This script compensates for the conjugation errors estimated by `neat_conj_est.py` by applying the inverse affine transformation H^−1\\hat{H}^{-1}H^−1 to an SLM pattern (generated as `slm_pattern.bmp` by `neat_learning.py`, which does not include conjugation error correction). The corrected pattern is saved as `k_vis_(args.dataset)_with_H.bmp`.\\
\\
> **Note:** If you are using a microscope with perfect conjugation and alignment, you can skip `neat_conj_est.py` and `neat_conj_corr.py`.\\
\\
* * *\\
\\
**Running the Capsule** \\
\\
This capsule demonstrates NeAT’s application on both custom-built and commercial microscopes:\\
\\
1. **Microscope with perfect conjugation and alignment**\\
\\
Run NeAT on a fixed mouse-brain-slice image stack (data in `/custom_built_brain_slice/`), where conjugation-error correction is not required.\\
\\
2. **Calibration for a microscope with possible conjugation and alignment issues (e.g., most commercial systems)**\\
\\
Process five calibration stacks, estimate conjugation errors, and apply those corrections to generate an error-corrected SLM pattern.\\
\\
3. **In vivo imaging with motion**\\
\\
Run NeAT on an in vivo mouse-brain image stack with sample motion. If the samples do not require motion correction (e.g., zebrafish larvaes or plants) but still need conjugation error correction, simply set `sample_motion = False` and proceed with this step.\\
\\
\\
Iksung Kang et al.](/content/explore/5c6c1dc5-7a81-4dc7-b337-ae4b2a533a3c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5265084/tree/v1/index.html)

published in [Nature Methods](/content/explore?query=Nature%20Methods&refine=journal/index.html)

[Social Sciences\\
\\
2 \| Mar \| 2026\\
\\
Ubiquitous Data-Driven Framework for Traffic Emission Estimation and Policy Evaluation\\
\\
This repository provides a full-stack urban transportation emissions analysis framework that integrates:\\
\\
1. Public camera feeds\\
2. Deep learning vehicle detection/classification\\
3. Simulation-based OD assignment\\
4. EPA MOVES Matrix emission modeling\\
5. Scenario testing for transportation policies (mode shift, peak spreading, congestion pricing)\\
\\
The framework is built on open datasets (NYC cameras, mobile OD, OpenStreetMap) and supports large-scale experiments.\\
\\
Songhua Hu](/content/explore/e669c49d-6622-45d5-95e4-b0554b24f2d1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0039779/tree/v1/index.html)

published in [Nature Sustainability](/content/explore?query=Nature%20Sustainability&refine=journal/index.html)

[Computer Science\\
\\
3 \| Apr \| 2026\\
\\
FAST-LoRa: An Efficient Simulation Framework for LoRaWAN Networks\\
\\
FAST-LoRa enables rapid, matrix-based simulation of large-scale LoRaWAN networks. It models interference, uplink transmissions, and gateway diversity to accurately estimate key metrics such as Packet Delivery Ratio and Energy Efficiency. The framework supports efficient transmission parameter optimization and ADR strategies, allowing fast evaluation and planning of dense IoT deployments without the overhead of packet-level simulations.\\
\\
Fabian Margreiter, Laura Acosta Garcia & Juan Aznar Poveda](/content/explore/d88bd8a1-e78f-4703-b4c4-1703e3603f87?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1001925/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/access.2026.3669347) published in [IEEE Access](/content/explore?query=IEEE%20Access&refine=journal/index.html)

[Computer Science\\
\\
1 \| Mar \| 2026\\
\\
Experimental data from Section III.C of the paper “A Robotic Laser Welding Seam Tracking Method for Complex 3-D Weld Trajectories Based on Geometric Structure Analysis” (DOI: 10.1109/TII.2026.3661009).\\
\\
Experimental data from Section III.C of the paper “A Robotic Laser Welding Seam Tracking Method for Complex 3-D Weld Trajectories Based on Geometric Structure Analysis” (DOI: 10.1109/TII.2026.3661009).\\
\\
Zhaoqi Chu et al.](/content/explore/5f2e8cef-4c7c-4c07-9353-272666d06f25?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1039143/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tii.2026.3661009) published in [IEEE Transactions on Industrial Informatics](/content/explore?query=IEEE%20Transactions%20on%20Industrial%20Informatics&refine=journal/index.html)

[Bioinformatics\\
\\
27 \| Feb \| 2026\\
\\
StrucGAP: a modular, streamlined and traceable data mining platform for structural and site-specific glycoproteomics\\
\\
StrucGAP is a modular toolkit for downstream analysis of glycoproteomics data. It includes support for preprocessing, glycan structure analysis, quantification, network visualization, functional annotation, and more.\\
\\
Shisheng Sun](/content/explore/236b3f40-2422-43ad-a312-fa3ee17282c7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8478302/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Earth Sciences\\
\\
27 \| Feb \| 2026\\
\\
Large-scale aggregation of humid heatwaves exacerbated by coastal oceanic warming\\
\\
The python and NCL codes for the article Cai et al. 2025 "Large-scale aggregation of humid heatwaves exacerbated by coastal oceanic warming".\\
\\
Fenying Cai](/content/explore/cf08f8f0-b927-43cc-a362-11b5b75f79a4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0627682/tree/v1/index.html)

published in [Nature Geoscience](/content/explore?query=Nature%20Geoscience&refine=journal/index.html)

[Earth Sciences\\
\\
27 \| Feb \| 2026\\
\\
Astronomical calibration of the middle Cambrian in Baltica: global carbon cycle synchronization and climate dynamics \| Supplementayr Text S2: Time Series Ptorocol\\
\\
Supplementary materila contianing the R code used to perform the Time series analysis of the following paper "Astronomical calibration of the middle Cambrian in Baltica: global carbon cycle synchronization and climate dynamics"\\
\\
Valentin Jamart et al.](/content/explore/ea7c43e1-c986-4ac5-97ad-d12032883ecc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1118709/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Biology\\
\\
26 \| Feb \| 2026\\
\\
Nuclear stiffening in neoplastic cells aggregates T cell exhaustion via pFAK/SP1/IL-6 axis in colorectal cancer\\
\\
Nuclear abnormalities such as nuclear deformation are hallmark of many diseases including cancer. Accumulating evidence suggests that the dense and mechanically stiff tumor microenvironment promotes nuclear deformation in cancer cells. However, little is known about how nuclear deformation in neoplastic cells regulates immune exhaustion in the tumor microenvironment. Here, we found that lamin A/C-mediated nuclear stiffening in neoplastic cells promotes nuclear translocation of phosphorylated focal adhesion kinase (pFAK), which was strongly correlated with the heterogeneity and exhaustion of CD8 + T cells within the spatial context of the tumor microenvironment in human colorectal cancer. Mechanistically, we reveal that increased nuclear tension within tumor cells promotes pFAK nuclear translocation, where nuclear pFAK was found to regulate SP1/IL-6-mediated T-cell exhaustion and the transcription of proinflammatory cytokines/chemokines. Pharmacological inhibition or disruption of pFAK nuclear translocation enhanced antitumor immune responses and synergistically potentiated αPD-1 and αTIM-3 immunotherapy by augmenting CD8 + T cell cytotoxicity and restoring exhaustion in preclinical models of colorectal cancer. These findings highlight the pivotal role of nuclear tension-mediated pFAK translocation into the tumor cell nucleus in regulating CD8 + T cell exhaustion, suggesting that pFAK is a promising target for advancing cancer immunotherapy.\\
\\
Hao Kong, Xiangji Wu, Qingxin Yang & Qian Sun](/content/explore/9cd8dd0e-3c26-4413-9c92-ce3573e5737d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7227472/tree/v1/index.html)

[Computer Science\\
\\
26 \| Feb \| 2026\\
\\
XA-Novo: An accurate and high-throughput mass spectrometry-based de novo sequencing technology for monoclonal antibodies and antibody mixtures\\
\\
source assembly develop code\\
update (20251106):\\
Discriminating Isoleucine from Leucine.\\
Coverage depth(7-mer).\\
Assembly path log (stdout.log).\\
\\
Wenbin Jiang](/content/explore/3810c1fa-1c20-4e9b-8606-06c1d00bbed1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5561269/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Biology\\
\\
20 \| May \| 2026\\
\\
maldipickr dereplicates microbial MALDI-TOF spectra to facilitate multiplexed isolation](/content/explore/5a879b70-3008-4e96-acd9-be0a817e9d46?page=2&filter=all/index.html)

[Summary: Microbiologists use MALDI-TOF for fast and cheap identification of microbial isolates. However, the standard procedure relies on a commercial database. Bioinformatic tools to dereplicate MALDI-TOF spectra have been developed, but an open and resource-efficient tool to reduce the redundancy of microbial isolates is lacking. Here we develop “maldipickr” for de novo-clustering of MALDI-TOF spectra to dereplicate and select isolates, and thereby facilitate large-scale cultivation projects.](/content/explore/5a879b70-3008-4e96-acd9-be0a817e9d46?page=2&filter=all/index.html)

[Availability and Implementation: The R package “maldipickr” is available through CRAN at](/content/explore/5a879b70-3008-4e96-acd9-be0a817e9d46?page=2&filter=all/index.html) [https://doi.org/10.32614/CRAN.package.maldipickr](https://doi.org/10.32614/CRAN.package.maldipickr) along extensive documentation at [https://clavellab.github.io/maldipickr](https://clavellab.github.io/maldipickr). Code and data to reproduce the analysis and figure are available at [https://github.com/ClavelLab/maldipickr\_manuscript](https://github.com/ClavelLab/maldipickr_manuscript) and [https://zenodo.org/records/15744631](https://zenodo.org/records/15744631).

Charlie Pauvert, David Wylensek, Selina Nüchtern & Thomas Clavel

[Open Capsule](/content/capsule/0812028/tree/v2/index.html)

[Bioinformatics\\
\\
3 \| Mar \| 2026\\
\\
High Recombination and Intra-oocyst Heterogeneity Shape Virulence in Cryptosporidium parvum, with CpDHR1 Defined as a Secondary Virulence Determinant\\
\\
Cryptosporidium parvum causes life-threatening diarrhea in infants and neonatal animals, with genetic recombination driving virulence variation, yet intra-oocyst genomic heterogeneity remains uncharacterized. Here, we crossed fluorescently tagged virulent IIdA20G1-HLJ and avirulent IIdA19G1-GD strains in IFN-γ knockout mice, performing whole-genome sequencing of single oocysts/sporozoites and establishing single-oocyst expansion lines. We found extreme intra-oocyst genomic heterozygosity, with sporozoites harboring mosaic parental alleles due to frequent crossovers. A high-resolution genetic map revealed an 8.3 kb/cM recombination rate with telomere-biased crossovers and 28 recombination hotspots near infectivity-associated genes. Expansion lines showed divergent virulence unlinked to gp60, and integrative comparative genomics identified cgd5\_4090 (CpDHR1), a DEAH-box RNA helicase, as a secondary virulence factor. Revsere genetic studies confirmed that CpDHR1 modulates parasite growth and virulence, possibly via effects on ribosome assembly due to conformational flexibility. These findings shed light on the evolution of C. parvum virulence and offer novel targets for cryptosporidiosis. control.\\
\\
Tianyi Hou](/content/explore/49f482e1-364a-4ae8-b7f4-7d5c64ae38c3?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0032926/tree/v2/index.html)

[Engineering\\
\\
25 \| Feb \| 2026\\
\\
A Foundational Generative Model for Breast Ultrasound Image Analysis](/content/explore/afc5eec6-ca0c-4ad8-8f2f-b9e26542e2f2?page=2&filter=all/index.html)

[This repository contains the official code of **A foundational generative model for breast ultrasound image analysis**. In this paper, we present **BUSGen**, the first foundational generative model specifically designed for breast ultrasound image analysis. Pretrained on over 3.5 million breast ultrasound images, BUSGen has acquired extensive knowledge of breast structures, pathological features, and clinical variations. With few-shot adaptation, BUSGen can generate repositories of realistic and informative task-specific data, facilitating the development of models for a wide range of downstream tasks. (1) BUSGen shows **high adaptivity** to significantly enhance breast cancer screening, diagnosis, and prognosis, even surpassing the performance of experienced radiologists. (2) Additionally, we characterized the **scaling effect** of generated data, and (3) showed that it improves the **generalization ability** of downstream models. (4) BUSGen enhances **privacy protection** by enabling fully de-identified data sharing, making progress forward in secure medical data utilization. An online demo of BUSGen is available at](/content/explore/afc5eec6-ca0c-4ad8-8f2f-b9e26542e2f2?page=2&filter=all/index.html) [https://aibus.bio](https://aibus.bio/).

Haojun Yu

[Open Capsule](/content/capsule/9938564/tree/v1/index.html)

published in [Nature Biomedical Engineering](/content/explore?query=Nature%20Biomedical%20Engineering&refine=journal/index.html)

[Medical Sciences\\
\\
24 \| Feb \| 2026\\
\\
Clinical and histopathological characterization of metastatic lobular breast cancer: lessons learned from post-mortem tissue donation programs\\
\\
While primary invasive lobular carcinoma (ILC) is well characterized, metastatic ILC remains understudied. Within the post-mortem tissue donation programs, UPTIDER (Belgium) and Hope for Others (USA), we first aimed to explore intra-patient heterogeneity of key prognostic and predictive markers (stromal tumor-infiltrating lymphocytes (sTIL), estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2) and KI67). Secondly, we compared detection of the metastases by pathology on autopsy samples versus pre-mortem imaging. In total, 306 metastases from 12 patients were collected at autopsy (median: 27 per patient). Both primary tumors (n = 15) and metastases (n = 232) had low sTIL levels, with a median of 2% (range: 0.67–6.67%) and 0.67% (range: 0–13.33%), respectively. Regression models showed lower ER- and PR-expression in metastases (respectively, n = 265 and n = 64) compared to primary tumors (both p < 0.01). KI67 was significantly higher in metastases (n = 262, p = 0.02). HER2-low metastases were found in all but one patient although in varying proportion of metastases (range: 7.5–100%). Central radiology and pathology review had a median concordance of 78% at organ level (range: 33.33–100%) and 71% (range: 55.88–85.29%) at patient level. Our findings suggest that a single metastatic biopsy has great limitations to guide treatment and that more adequate methods are needed to detect and monitor ILC metastases.\\
\\
Gitte Zels et al.](/content/explore/42e5bcd4-a9ef-44b0-b5f3-bf2d2fbb90ce?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1000635/tree/v1/index.html)

[Associated article](https://doi.org/10.1038/s41523-026-00912-5) published in [npj Breast Cancer](/content/explore?query=npj%20Breast%20Cancer&refine=journal/index.html)

[Earth Sciences\\
\\
24 \| Feb \| 2026\\
\\
Fresh, salty, and still not talking?\\
\\
This capsule contains the fully reproducible R scripts that generate all figures for the manuscript "Fresh, salty, and still not talking? Assessing and addressing the limnology-oceanography divide" (Díaz-Torres et al., in preparation). The study combines a bilingual global survey of aquatic researchers (n = 103 from 27 countries) and a synoptic literature review of papers published in Limnology & Oceanography (n = 319, 2015–2025) to assess the collaboration divide between limnology and oceanography. The scripts produce five main figures: survey overview (world map, organization types, research areas), complexity analysis of research areas and language distribution, collaboration frequency, gaps analysis (general and by language), and barriers and modeling priorities (four‑panel figure). All figures are rendered in publication‑ready quality using consistent color palettes and Arial font, ensuring exact reproducibility of the visual results.\\
\\
Osiris Díaz Torres et al.](/content/explore/2232ba36-3ba6-4b06-ac5b-c0a09dc3ae27?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6693354/tree/v1/index.html)

[Engineering\\
\\
23 \| Feb \| 2026\\
\\
An Improved Frequency Estimator Using Auxiliary DFT Samples: CRB Performance Across All Signal Lengths\\
\\
An ultra-precision, fast, and accurate frequency estimation algorithm. See: A. Serbes, "An Improved Frequency Estimator Using Auxiliary DFT Samples: CRB Performance Across All Signal Lengths," in IEEE Transactions on Aerospace and Electronic Systems, vol. 61, no. 6, pp. 18970-18981, Dec. 2025.\\
\\
Ahmet Serbes](/content/explore/b9f5d6b8-eabb-4980-befd-f6050867acd2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4586312/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/taes.2025.3617927) published in [IEEE Transactions on Aerospace and Electronic Systems](/content/explore?query=IEEE%20Transactions%20on%20Aerospace%20and%20Electronic%20Systems&refine=journal/index.html)

[Computer Science\\
\\
23 \| Feb \| 2026\\
\\
Adaptive Basis Functions for Enhanced Function Approximation in Kolmogorov–Arnold Networks\\
\\
This paper proposes adaptive basis functions (AdaKANs) for Kolmogorov–Arnold Networks (KANs), a novel addition to KANs that enables optimized adjustment of univariate basis functions during training. KANs, with additional Adaptive Basis Functions (AdaKANs) achieve smoother convergence, improved generalization, and an average reduction in mean-squared error by 50% on noisy polynomial functions, and 3 benchmark datasets compared to a regular KAN, Fourier Function Networks, and a regular MLP. This method addresses the limitations of static basis functions, further improving the scalability, flexibility, and applicability of KANs in complex function approximation tasks.\\
\\
Aditya Chakraborty](/content/explore/ed9e3b2c-f006-48af-a23e-0d4bd75b8e1c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3608043/tree/v1/index.html)

[Engineering\\
\\
2 \| Mar \| 2026\\
\\
Density-Driven Optimal Control for Efficient and Collaborative Multiagent Nonuniform Coverage\\
\\
An implementation of Optimal Transport-based multi-robot/multi-agent nonuniform area coverage. This framework achieves efficient and collaborative density-driven multi-agent coordination in a decentralized manner.\\
\\
Kooktae Lee](/content/explore/ebd9df94-0989-426b-9930-b4020a3920f3?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7735062/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tsmc.2025.3622075) published in [IEEE Transactions on Systems, Man, and Cybernetics: Systems](/content/explore?query=IEEE%20Transactions%20on%20Systems%2C%20Man%2C%20and%20Cybernetics%3A%20Systems&refine=journal/index.html)

[Social Sciences\\
\\
23 \| Feb \| 2026\\
\\
Premature termination of unemployment benefits increased COVID-19 transmission and deaths in the USA\\
\\
The code and data in this replication package construct the analysis file for "Premature termination of unemployment benefits increased COVID-19 transmission and deaths in the USA" from the 26 data sources using Stata 19-MP. The `main.do` file runs all of the code to generate the data for the two tables and six figures in the main document, and 50 exhibits in the Appendix. The replicator should expect the manuscript exhibits code (`99.02_main_exhibits.do`) to run for 6 minutes. The code for raw file to generated data `99.01_main_raw_prep.do` takes 15 minutes to run, and the code for Supplementary Materials `99.03_supplementary_exhibits.do` takes 1 hour 30 minutes to run.\\
\\
Sungbin Park, Kyung Min Lee & John S Earle](/content/explore/8e9d7a12-7240-458e-a90f-62c9bcdb67cf?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0776661/tree/v1/index.html)

published in [Nature Human Behaviour](/content/explore?query=Nature%20Human%20Behaviour&refine=journal/index.html)

[Engineering\\
\\
23 \| Feb \| 2026\\
\\
A Comprehensive Framework for Predictive Maintenance in Building Facilities Using Machine Learning and Environmental Triggers\\
\\
This research dashboard delivers end-to-end analytical and predictive capabilities for facility management and building engineering. It transforms raw maintenance logs into actionable insights, enabling data-driven decisions that improve reliability, reduce costs, and optimize resource allocation.\\
\\
**Core Capabilities** \\
\\
**📊 Exploratory Data Analysis (EDA)**\\
\\
- Identify historical trends in work orders and maintenance activities\\
- Compare cost implications of proactive versus reactive maintenance strategies\\
- Analyze how building condition (Facility Condition Index, FCI) and weather patterns influence HVAC and system failures\\
\\
**🤖 Machine Learning Integration**\\
\\
- **Cost Prediction**: \\
Estimate total work order costs using factors such as building age, system type, and temperature conditions.\\
- **Delay Risk Assessment**: \\
Predict the likelihood of significant delays based on component type, labor availability, and operational constraints.\\
- **Building Archetype Clustering**: \\
Group facilities into distinct clusters according to size, condition, and maintenance frequency to support strategic planning and optimized resource deployment.\\
\\
Rahat Aayaz](/content/explore/a570e3c3-020e-441f-ba83-c36461fd0184?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2746458/tree/v2/index.html)

[Computer Science\\
\\
24 \| Feb \| 2026\\
\\
POIROT (PrOfIling RObotics Tool)\\
\\
POIROT (PrOfIling RObotics Tool) is a comprehensive profiling and monitoring framework for ROS 2 applications that provides detailed insights into function performance, resource consumption, energy usage and CO2 emissions.\\
\\
Miguel Á González-Santamarta, Sergio Sánchez de la Fuente & Francisco J. Rodríguez-Lera](/content/explore/946bf625-6290-4486-9897-0e1bb543cacb?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1513431/tree/v1/index.html)

[Engineering\\
\\
21 \| Feb \| 2026\\
\\
Omnidirectional Imaging FMCW Radar System for Rescue Robotics\\
\\
A compact real aperture omnidirectional imaging radar system for use in robotic search and rescue operations is presented. This system uses an ultra-wideband frequency-modulated continuous wave (FMCW) radar sensor operating at 80 GHz and a mechanically steered lens-mirror antenna configuration to capture images comparable to 360-degree photographs. The antenna configuration is optimized for a compact footprint on a mobile robot and a small moment of inertia for fast scans. The directivity of the lens-mirror arrangement is characterized, and the influence of the mirror dimensions on side lobes is investigated. This reveals spillover effects that are inversely proportional to mirror size. Key factors that influence the scan time are investigated, including the chirp repetition rate of the radar sensor and the time required to change the direction of the radar beam. The investigation shows a compromise between FMCW bandwidth and scan time. A signal processing method that addresses the processing steps from input samples to output images is presented, with a focus on the detection process. This process uses range and image data to determine the detection threshold. The imaging system is applied in a rescue robotics scenario to demonstrate its ability to visualize human shapes and complex indoor environments with scan times of up to two minutes.\\
This capsule contains code to perform the signal processing methods on the data captured with the radar system.\\
\\
Marc Hamme, Leon Hülsmann, Tobias T. Braun & Nils Pohl](/content/explore/2ba62b91-aa07-45d1-b1bf-fc3d62dde52d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8486865/tree/v1/index.html)

[Computer Science\\
\\
21 \| Feb \| 2026\\
\\
Discrete Solution Operator Learning (DiSOL)\\
\\
This Code Ocean capsule provides an executable implementation of Discrete Solution Operator Learning (DiSOL) for geometry-dependent PDE operator learning, accompanying the manuscript: Discrete Solution Operator Learning for Geometry-Dependent PDEs. The capsule is designed to (i) demonstrate that the codebase is runnable end-to-end, (ii) reproduce the basic training/validation/testing pipeline for all benchmark problems, and (iii) provide example outputs and checkpoints under the compute/storage constraints of the Code Ocean environment.\\
\\
Jinshuai Bai & Haolin Li](/content/explore/e9995e58-6571-4f81-adb6-2de110cf8704?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4396603/tree/v1/index.html)

[Biology\\
\\
21 \| Feb \| 2026\\
\\
Copy number signatures in cervical samples enable early detection of high-grade serous ovarian carcinoma\\
\\
Ovarian cancer is often diagnosed in advanced stages, resulting in poor outcomes. There is an unmet need for a sensitive and specific screening tool for early-stage detection of ovarian cancer. Recognizing that high-grade serous ovarian carcinoma (HGSC) is driven by copy number alterations (CNAs) and that tumor DNA can be detected in cervical samples, we analyzed CNAs from shallow whole genome sequencing of 212 cervical samples from 128 women with/without HGSC, including 30 germline BRCA1/2 mutation carriers. Using a machine-learning classifier, we developed High-grade serous ovarian cancer Cervical copy number signature (HCsig), a predictor for HGSC detection. HGSC-derived CNAs were detectable with HCsig in cervical samples collected several years before diagnosis. HCsig correctly identified HGSC in 79% of archival cervical samples, importantly including 91% stage I-II (0-27 months pre-diagnosis), and 77% stage III-IV (0-65 months pre-diagnosis). Among patients with HGSC who had multiple pre-diagnostic samples collected during the pre-symptomatic phase, 85% had at least one HCpositive cervical sample before surgery (up to 65 months before diagnosis). Detection rates were 82% and 76% for BRCA1/2-mutated and wildtype HGSC, respectively. Validation in 172 independent samples (0-98 months pre-diagnosis) showed 76% sensitivity and 94% specificity (AUC=0.83), including high sensitivity for early-stage cancers. We show that applying the HCsig classifier to cervical samples, including from non-symptomatic women several years before diagnosis, holds promise for early-stage detection and secondary prevention of HGSC. Moreover, it may serve as a screening tool to aid decision-making regarding timing of risk-reducing surgery in high-risk populations.\\
\\
Srinivas Veerla](/content/explore/ae500d95-13e1-449c-9807-41ac9482ec44?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7890879/tree/v1/index.html)

[Biology\\
\\
20 \| Feb \| 2026\\
\\
DNA damage drives antigen diversification through mosaic Variant Surface Glycoprotein (VSG) formation in Trypanosoma brucei\_VSG-AMP-SeqPipeline\\
\\
This pipeline takes VSG-AMP-Seq reads and processes them to identify mosaic VSGs based upon a Target VSG of interest.\\
\\
Monica Mugnier & Jaclyn Smith](/content/explore/6b48447e-a1ef-40bb-a00d-a46e9fac9e2e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5966016/tree/v1/index.html)

published in [Nature](/content/explore?query=Nature&refine=journal/index.html)

[Biology\\
\\
20 \| Feb \| 2026\\
\\
DNA damage drives antigen diversification through mosaic Variant Surface Glycoprotein (VSG) formation in Trypanosoma brucei\_R-scripts-mosaic-VSG-graphing\\
\\
This R instance takes the output from VSG-AMP-Seq, imports the data into R, and generates graphs.\\
\\
Monica Mugnier & Jaclyn Smith](/content/explore/8a7d2faf-1f1e-4999-97df-68631acbcb21?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2066134/tree/v1/index.html)

published in [Nature](/content/explore?query=Nature&refine=journal/index.html)

[Medical Sciences\\
\\
20 \| Feb \| 2026\\
\\
The jagged edge of ChatGPT Health: Under-triage in consumer-facing artificial intelligence\_R1\\
\\
ChatGPT Health launched in January 2026 as OpenAI’s consumer health tool, reaching millions of users. Here, we conducted a structured stress test of triage recommendations using 60 clinician-authored vignettes across 21 clinical domains under 16 factorial conditions (960 total responses). Performance followed an inverted U-shaped pattern, with the most dangerous failures concentrated at clinical extremes: non-urgent presentations (35%) and emergency conditions (48%). Among gold-standard emergencies, the system under-triaged 52% of cases, directing patients with diabetic ketoacidosis and impending respiratory failure to 24–48-hour evaluation rather than the emergency department, while correctly triaging classical emergencies such as stroke and anaphylaxis. When family or friends minimized symptoms (anchoring bias), triage recommendations shifted significantly in edge cases (OR 11.7, 95% CI 3.7-36.6), with the majority of shifts toward less urgent care. Crisis intervention messages activated unpredictably across suicidal ideation presentations, firing more when patients described no specific method than when they did. Patient race, gender, and barriers to care showed no significant effects, though confidence intervals did not exclude clinically meaningful differences. Our findings identify two distinct safety failures – selective under-triage of evolving emergencies and unpredictable crisis guardrail activation – underscoring the need for independent safety evaluation before consumer artificial intelligence deployment at scale.\\
\\
Ashwin Ramaswamy](/content/explore/910ba3ae-418b-436b-8d30-559649514500?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7535456/tree/v1/index.html)

published in [Nature Medicine](/content/explore?query=Nature%20Medicine&refine=journal/index.html)

[Medical Sciences\\
\\
20 \| Feb \| 2026\\
\\
The jagged edge of ChatGPT Health: Under-triage in consumer-facing artificial intelligence\\
\\
Re: Manuscript NMED-FT148925\\
\\
Ashwin Ramaswamy](/content/explore/c80218b4-1787-4d99-bbce-7cc73b56cdea?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3063310/tree/v1/index.html)

published in [Nature Medicine](/content/explore?query=Nature%20Medicine&refine=journal/index.html)

[Chemistry\\
\\
20 \| Feb \| 2026\\
\\
Single-Residue Solvation Perturbations Regulate Global Protein Architecture and Function\\
\\
Protein-water interactions fundamentally shape the structure, stability, dynamics, and functionality of proteins. However, the heterogeneous nature of the protein-water interface and the disparity in their dynamic interplay make it challenging to understand how local water perturbations influence protein structural dynamics over space and time. In this study, we introduce a photochromic molecule, spiropyran, to modify a specific residue of proteins, thereby achieving a reversible, residue-specific, and amplified perturbation on the hydrophobicity of protein surfaces. With the aid of controlled, amplified hydrophobic perturbations, we reveal that even subtle, residue-level changes in hydrophobicity induce significant global alterations in protein hydration patterns. These hydration shifts propagate in an amino acid sequence-dependent manner, initiating a “butterfly effect” that dramatically influences overall protein architecture and catalytic performance. Our findings establish that interfacial water networks not only capture the surface physicochemical patterns of proteins but also mediate the propagation of local perturbations into broader structural and functional fluctuations. By shifting the paradigm from “structure-function” to “structure-hydration-function”, our work provides innovative perspectives into understanding protein architecture and guiding future drug design strategies.\\
\\
Yingya Liu et al.](/content/explore/533bfdbe-d1d6-4713-b77a-fa8eea73b9cc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1433564/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Medical Sciences\\
\\
19 \| Feb \| 2026\\
\\
Serglycin Drives LAG3+ Treg Differentiation and Immunosuppression in Gastric Cancer\\
\\
Integrated single-cell sequencing data from 23 pairs of gastric cancer and corresponding normal gastric mucosa, isolated T cells and performed further analysis.\\
\\
Qingyuan Wang](/content/explore/a2f27871-5c86-4cca-a3a8-168f8d7b4cee?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1513246/tree/v1/index.html)

[Mathematics\\
\\
19 \| Feb \| 2026\\
\\
octAFEM - Numerical investigation of an adaptive least-squares finite element method for the solution of Zarantonello-linearized first-order systems of quasilinear PDEs\\
\\
This is the octAFEM package for the solution of first-order systems of quasi-linear PDEs using Zarantonello linearization with lowest-order least-squares finite element discretization. The Zarantonello fixed-point iteration is an established linearization scheme for quasilinear PDEs with strongly monotone and Lipschitz continuous nonlinearity in Hilbert spaces. This software investigates a weighted least-squares minimization for the computation of the update of this scheme. The resulting formulation allows for a conforming least-squares finite element discretization of the primal and dual variable of the PDE. The least-squares functional provides a built-in a posteriori discretization error estimator in each linearization step motivating an adaptive Uzawa-type algorithm with an outer linearization loop and an inner adaptive mesh-refinement loop. The numerical experiments illustrate the performance of the algorithm. Particular focus is on the role of the weights in the least-squares functional of the linearized problem and their influence on the robustness of the Zarantonello damping parameter.\\
\\
The code is tested and developed under MATLAB version 25.1.0.2973910 (R2025a) Update 1. It should be executable with older versions as well. The main features include\\
\\
- full Octave compatibility (tested with version 10.3.0)\\
- improved performance by vectorization\\
- parallelization of discretization\\
- object-oriented realization of adaptive mesh-refinement with possible separate marking for data approximation\\
\\
Philipp Bringmann & Dirk Praetorius](/content/explore/97d736f8-bf88-4183-8799-0eacfa78fba0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4427783/tree/v1/index.html)

[Engineering\\
\\
7 \| May \| 2026\\
\\
Performance of Sequential And Quickest Detection Based on Machine Learning\\
\\
**Abstract:** With the rapid proliferation of Machine-Learning (ML) and Deep Learning (DL) based decision systems, properly characterizing their often unpredictable performance is a key challenge. In this work we introduce data-driven analogues to the Sequential Probability Ratio Test (SPRT), and Page Test, denoted the Sequential Data-Driven Decision Function (S-D3F) and Quickest Data-Driven Decision Function (Q-D3F), respectively. It is shown that the S-D3F and Q-D3F inherit many key properties from the SPRT and Page test which allow for a reliable extension of their performance prediction strategies to these data-driven approaches via the process of statistical characterization. As a result, one is able to perform sequential analysis and quickest detection using a ML driven detector while attaining near-optimal performance and reliable performance modeling; that is, the performance can be predicted from the training data itself. The reliability of these predictions is demonstrated via extensive numerical experimentation, including a challenging low-SNR passive acoustic detection problem with an ambient acoustic dataset injected with a realistic synthetic signal.\\
\\
**Expected Duration of Reproducible Run:** ~2 hours\\
\\
Ryan J. Harvey, Paolo Braca, Leonardo M. Millefiori & Peter K. Willett](/content/explore/63890357-b569-4a38-ac05-c1b461aa9d05?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5227518/tree/v2/index.html)

[Engineering\\
\\
17 \| Mar \| 2026\\
\\
Explainable AI reveals modeling can outweigh aging in infrastructure resilience\\
\\
Escalating climate hazards are amplifying the vulnerability of aging infrastructure, requiring robust, cross-system resilience assessments. Traditional resilience models are computationally intensive, often limiting scalability and timely application, while machine learning (ML) approaches often lack interpretability, undermining trust in operational and policy-level decision-making. We argue for a transformational departure from traditional resilience assessment and opaque machine-learning models by introducing NestedSHAP. This is a first-of-its-kind explainable AI (XAI) framework that combines conditional Shapley values, hierarchical clustering, and representative-instance selection to efficiently resolve feature interactions in models predicting the flood resilience of more than 100,000 models of aging bridges. NestedSHAP corrects biases inherent in marginal Shapley values, which implicitly assume feature independence. The proposed model provides more reliable, interaction-aware insights at a fraction of the computational cost compared to full conditional Shapley implementations. Accordingly, we develop a methodological framework that translates complex model predictions into transparent drivers of resilience loss and recovery. The framework enables policymakers and infrastructure owners to prioritize adaptation strategies under climate and budget constraints. Our findings are inherently thought-provoking: methodological accuracy can dominate over the impact of asset age in resilience assessment, and reliable methods must be enforced in standards for infrastructure design and management.\\
\\
Ali Amini & Azam Abdollahi](/content/explore/048332a5-9e40-463e-9622-e6aaf99e368a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6690106/tree/v1/index.html)

[Engineering\\
\\
17 \| Feb \| 2026\\
\\
Calculation of Positive-Sequence and Zero-Sequence Impedances of Three-Phase Submarine Cables\\
\\
This code describes semi-analytical calculation methods for the positive-sequence and zero-sequence impedances of 3-phase armored cables with solid sheaths and a single armor layer. Separate equivalent circuits are defined for these two impedances. Subsequently, circuit component values are calculated based on 2D field equations, with some 3D effects modeled. The existing approach for calculating zero-sequence impedance is corrected and expanded by inclusion of additional inductances due to solenoidal fields and magnetic armor effects. 3D finite-element models of cables with magnetic and non magnetic armor are used to assess the accuracy of our circuit models; discrepancies are less than 3% at frequencies between 50 Hz and 5 kHz. Python code with the implementation of the method is provided for accessibility.\\
\\
Kevin Goddard et al.](/content/explore/245f45ba-9b25-44ef-8f91-251e2ca5fa46?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3085584/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tpwrd.2026.3657673) published in [IEEE Transactions on Power Delivery](/content/explore?query=IEEE%20Transactions%20on%20Power%20Delivery&refine=journal/index.html)

[Earth Sciences\\
\\
16 \| Feb \| 2026\\
\\
Codes for Declining anthropogenic aerosols amplify Northern Hemisphere Hadley circulation weakening in the 21st century\\
\\
Scripts and data for the paper about responses of the Hadley circulation to aerosol changes.\\
\\
Seo-Yeon Kim](/content/explore/df4123cf-30ff-4656-92e0-bd36e8fe73b4?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2041858/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Computer Science\\
\\
14 \| Feb \| 2026\\
\\
A Trainable-Embedding Quantum Physics-Informed Framework for Multi-Species Reaction–Diffusion Systems\\
\\
Supplementary code\\
\\
Ban Tran, Nahid Binandeh Dehaghani, A. Pedro Aguiar, Rafal Wisniewski, Susan Mengel](/content/explore/2943bf13-5b21-47e9-bcbe-0dc8a56a654a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8043889/tree/v1/index.html)

[Social Sciences\\
\\
14 \| Feb \| 2026\\
\\
Spatial perspective on environmental migration: Empirical insights from a spatio-temporal approach in the United States, 1970–2010\\
\\
This capsule contains data and code for the paper titled "Spatial perspective on environmental migration: Empirical insights from a spatio-temporal approach in the United States, 1970–2010"\\
\\
Shuai Zhou, Guangqing Chi & Chuan Liao](/content/explore/4aba7384-d5c3-43a8-9fa9-9e4c56194a97?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7881931/tree/v1/index.html)

[Engineering\\
\\
14 \| Feb \| 2026\\
\\
Contrast-based GED cleaning for ambulatory EEG\\
\\
Motion artifacts in EEG recordings, whether from subject movement, electrode displacement, or environmental interference, introduce noise that can obscure meaningful brain activity. This project focuses on developing an unsupervised method for decomposing multichannel EEG to isolate large amplitude artifacts such as motion.\\
\\
Sahar Sattari & Lyndia Wu](/content/explore/a708d0c1-09d2-468a-8269-5d35eb85ff0c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3394951/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tbme.2025.3630112) published in [IEEE Transactions on Biomedical Engineering](/content/explore?query=IEEE%20Transactions%20on%20Biomedical%20Engineering&refine=journal/index.html)

[Biology\\
\\
14 \| Feb \| 2026\\
\\
eScreen: a deep learning framework for functionally decoding the regulatory genome at single-nucleotide resolution](/content/explore/99d9abcf-d00d-4436-83b5-b5ff85f4acdb?page=2&filter=all/index.html)

[The human genome is densely populated with cis-regulatory elements (CREs), yet deciphering their functional regulatory syntax and combinational logic remains a fundamental challenge. Here, we integrate 379 genome-scale CRISPR screen experiments, encompassing 21 million perturbations across 23 cell types, to construct a compendium of 41,239 high-confidence functional CREs from 530,527 candidates. Leveraging this resource, we develop eScreen, a deep learning model built on the StripedHyena2 architecture to functionally decode the regulatory genome at single-nucleotide resolution. eScreen achieves three primary functions: (1) predicts genome-wide cell-type-specific CRE functional activity with high accuracy, outperforming existing models; (2) provides mechanistic interpretation of regulatory syntax at single- nucleotide resolution; (3) dissects the functional organization of enhancer clusters through in silico perturbation analysis. We perform multiple independent CRISPR knockout, CRISPR interference (CRISPRi), and base editing screens to validate these functions of eScreen both at scale and on individual cases. Furthermore, we provide an interactive web server (](/content/explore/99d9abcf-d00d-4436-83b5-b5ff85f4acdb?page=2&filter=all/index.html) [https://escreen.huanglabxmu.com/](https://escreen.huanglabxmu.com/)) for the community to access the integrated CRISPR screen resources and eScreen functions. Collectively, our work establishes a highly precise and convenient tool to decode the causal effects of the regulatory genome.

Shijie Luo et al.

[Open Capsule](/content/capsule/1490279/tree/v1/index.html)

[Computer Science\\
\\
13 \| Feb \| 2026\\
\\
Reliable uncertainty estimates in deep learning with efficient Metropolis-Hastings algorithms\\
\\
The code to run the sampling experiments to reproduce the paper 'Reliable uncertainty estimates in deep learning with efficient Metropolis-Hastings algorithms', currently in review to Nature Communications.\\
\\
Matthias Schmal](/content/explore/c586e748-168e-4f38-a381-98225813c7c5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7879717/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Computer Science\\
\\
12 \| Feb \| 2026\\
\\
Automated Phenotyping of Faba Bean Seeds Using Segment Anything 2 for Trait Extraction\\
\\
This capsule implements a fully automated, training-free workflow for extracting phenotypic traits from bean seed images. The pipeline uses the pretrained Segment Anything Model (SAM2.1, Meta AI) to perform high-precision instance segmentation without manual prompts or annotated training data, followed by custom post-processing for trait quantification.\\
The workflow extracts calibrated morphological features (size, shape, and spatial measurements) from binary seed masks using a reference coin and performs colorimetric analysis using a 24-patch color card to characterize dominant seed coat color. Thousand grain weight (TGW, g) is also included in the pipeline. \\
For computational efficiency, the capsule demonstrates the complete pipeline using a small representative subset of images. Full-scale analyses reported in the manuscript were conducted using the same workflow.\\
\\
Harpreet Kaur Bargota et al.](/content/explore/cc0659f5-9e4d-46cb-839c-883a2db286fe?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4782556/tree/v1/index.html)

[Computer Science\\
\\
12 \| Feb \| 2026\\
\\
Skip-VFL: Mitigating Extreme Class Imbalance in Feature-Heterogeneous Vertical Federated Learning for Intrusion Detection\\
\\
Distributed network environments increasingly rely\\
on Vertical Federated Learning (VFL) to address the growing\\
heterogeneity of device features, which conventional Federated\\
Learning is ill-equipped to handle. While VFL facilitates collab-\\
oration across diverse network entities in detecting anomalies,\\
current models often remain impractical for long-tailed datasets,\\
especially in network security domain, as they fail to account for\\
the inherent class imbalances that bias models toward majority\\
classes.\\
Therefore, We propose Skip-VFL, a novel tiered framework\\
for robust anomaly detection in heterogeneous networks. Skip-\\
VFL organizes data into three categories: shared core fea-\\
tures (Fcore), overlapping features (Foverlap), and unique features\\
(Funique), bridging the gap between resource-constrained IoT\\
devices and high-capacity servers. To address class imbalances,\\
Skip-VFL introduces a two-stage training approach: Latent\\
Space Alignment, which clusters feature representations, and\\
Weighted Classification using Inverse Class Frequency (ICF)\\
to increase sensitivity to rare anomalies. A key innovation is\\
the Skip-Layer Mechanism, which preserves feature privacy by\\
withholding tier-specific adapters (ϕk ) from the central server\\
while enabling global model synchronization.\\
Experiments on NSL-KDD, UNSW-NB15, and ToN-IoT\\
datasets show that Skip-VFL outperforms baseline methods,\\
achieving 88.4% recall on NSL-KDD in extreme 1% imbal-\\
ance scenarios. Unlike peers, which struggles with instability,\\
Skip-VFL maintains consistent performance, proving it to be\\
a reliable, privacy-preserving solution for anomaly detection in\\
decentralized networks.\\
\\
Alimov Abdulboriy Abdulkhay ugli & Ji Sun Shin](/content/explore/0d00a4de-3ad2-492c-ba09-b9d37e1839dc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8042880/tree/v1/index.html)

[Physics\\
\\
12 \| Feb \| 2026\\
\\
Isotopic shift on the spectra of photonic meta-atoms (code and example)\\
\\
This Code Ocean capsule provides an example demonstrating the calculation of a photonic meta-atom for a higher-order nonlinear Schrödinger equation via a split-step Fourier method. It also provides tools to perform a phase-matching analysis to predict resonance loci of radiation modes, shed by the meta-atom.\\
\\
Oliver Melchert](/content/explore/1b25c554-d283-4ba2-93d3-b9101aed5186?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9247289/tree/v1/index.html)

[Chemistry\\
\\
25 \| Feb \| 2026\\
\\
Single-cell thiol profiling enabled by live-cell labeling reveals metabolic heterogeneity in ferroptosis\\
\\
The untargeted single-cell metabolome library was established based on UPLC-MS/MS identification at population level and algorithmic recognition at the single-cell level. The single-cell metabolome information could be extracted by a four-step workflow based on algorithm: format transformation, EIC extraction, single-cell event authentication and normalization.\\
\\
Daiyu Miao](/content/explore/7d166673-c111-4d51-ba13-eee417449a6e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1549682/tree/v3/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
11 \| Feb \| 2026\\
\\
Neurovascular Impulse Response Function (IRF) during spontaneous activity differentially reflects intrinsic neuromodulation across cortical regions\\
\\
Ascending neuromodulatory projections from deep brain nuclei generate internal brain states that differentially engage specific neuronal cell types. Because neurovascular coupling is cell-type specific and neuromodulatory transmitters have vasoactive properties, we hypothesized that the impulse response function (IRF) linking spontaneous neuronal activity with hemodynamics would depend on neuromodulation. Here, we use widefield cortical imaging to observe the resting state relationship between population level neuronal Ca2+ activity, fluctuations in oxygenation and concentration of hemoglobin, and release of the vasoactive neuromodulators Norepinephrine (NE) and Acetylcholine (ACh). First, the IRF linking neuronal activity and the hemodynamic response failed to predict hemodynamic fluctuations during periods marked by higher arousal (high NE and pupil diameter). Second, hemodynamic fluctuations were well predicted by a regression model factoring in both Ca2+ activity and NE release. Third, Ca2+ and hemodynamic functional connectivity patterns diverged during periods of high arousal. Without accounting for NE neuromodulation and the associated vasoconstriction, diminished hemodynamic coherence, commonly referred to as “functional (dys)connectivity” in BOLD fMRI studies, can be falsely interpreted as neuronal desynchronizations.\\
\\
Rauscher, Bradley et al.](/content/explore/7b062934-99da-412c-8c47-b1f98d8d5f91?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9422478/tree/v1/index.html)

published in [Nature Neuroscience](/content/explore?query=Nature%20Neuroscience&refine=journal/index.html)

[Earth Sciences\\
\\
11 \| Feb \| 2026\\
\\
Increased Spread of Global Flash Droughts Threatens Vegetation Productivity Resilience\\
\\
Flash drought has become increasingly common. We used multi-source data to find the different responses of vegetation productivity to flash and slow droughts and analyzed the potential drivers.\\
\\
Renjie Guo](/content/explore/8ef271af-bf36-4bc2-92de-051db3403e63?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0553694/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Physics\\
\\
11 \| Feb \| 2026\\
\\
Optically Detected Nuclear Magnetic Resonance Capsule: Data and Programs\\
\\
Abstract: Nuclear magnetic resonance (NMR) is a powerful tool for applications ranging from chemical analysis to quantum information processing. Achieving optical initialization and detection of molecular nuclear spins promises new opportunities — including improved NMR signals at low magnetic field, sensitivity down to the single-molecule level, and full access to atomically precise molecular architectures for quantum technologies. In this study, we report optical readout of coherently controlled nuclear spins in a europium-based molecular crystal. By harnessing ultra-narrow optical transitions, we achieve optical initialization and detection of nuclear spin states. Through radio-frequency driving, we address two nuclear quadrupole resonances, characterized by narrow inhomogeneous linewidths and a distinct correlation with the optical transition frequency. We implement Rabi oscillations, spin echo and dynamical decoupling techniques, achieving nuclear spin quantum coherence with a lifetime of up to 2 ms. These results highlight the capabilities of optically detected NMR (ODNMR) and underscore the potential of molecular nuclear spins for quantum information processing.\\
\\
In this capsule the data and programs are uploaded for the review process: \\
\\
- each plot in the manuscript was created with an python plot program, the corresponding data is uploaded in the data folders\\
- Supplementary Information (SI) plots have their programs and folders\\
- all plots created by the programs are saved in the results folder\\
- more information can be found in the readme.txt documentation file\\
\\
Evgenij Vasilenko et al.](/content/explore/2ab33539-bc21-469a-93db-aff8bd5659d7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0050043/tree/v1/index.html)

published in [Nature Materials](/content/explore?query=Nature%20Materials&refine=journal/index.html)

[Medical Sciences\\
\\
11 \| Feb \| 2026\\
\\
Code for Quality of childbirth care provided by skilled birth attendants: A conceptual framework, new estimate of effective coverage and indicator assessment in Nepal, Senegal, and Zambia\\
\\
From manuscript abstract: Globally, recent increases in skilled birth attendant (SBA) coverage do not appear to be translating to proportionate reductions in maternal and neonatal mortality. One hypothesized driver is a quality gap in the childbirth care provided by health professionals. This research was conducted as part of a multi-country study to assess the quality of pre-service education of health professionals providing childbirth care and to enhance the measurement related to the coverage of SBAs and SHP. This work introduces a new metric of SBA quality, uses that metric to identify tracer indicators of quality that could be integrated into routine monitoring without the use of complex statistical methods, and identifies priorities for improving quality of care in Nepal, Senegal and Zambia.\\
\\
We gathered primary and secondary data, aligned to a comprehensive conceptual framework about quality of maternal and newborn care, in three “Exemplar” countries (Nepal, Senegal and Zambia). We applied latent variable analysis principles to measure quality, and constructed estimates of effective coverage of SBA. \\
\\
The five highest weighted indicators most predictive of quality of care were presence of a national compulsory continued development system, presence of a national licensing, relicensing and registration system, presence of a national education regulation system, proportion of health facilities with electricity available at all times, and proportion of health facilities with amoxicillin suspension available. Validity tests showed that effective SBA coverage had a stronger relationship with early neonatal mortality rates than crude SBA coverage at the subnational level. Our analysis suggests that effective SBA coverage is substantially lower than crude SBA coverage, ranging from 18% to 30% in Senegal, 25% to 43% in Nepal, and 12% to 35% in Zambia.\\
\\
Lower effective coverage levels demonstrated here reveal that mothers and newborns are receiving far less of the essential care than official statistics may portray. This work presents a novel approach to measure one of the most challenging concepts in global health, and the highest weighted indicators in this analysis may be useful for routine monitoring in LMICs and at the global level. These results also add to the evidence supporting a basic policy package to improve the quality of maternal and newborn care, including a national compulsory continued education system, national licensing, relicensing and registration systems and national education regulation systems. Finally, programs in these countries may benefit from using the subnational estimates to geographically tailor interventions and to improve health outcomes.\\
\\
David Phillips, Yao He & Laith Hussain-Alkhateeb](/content/explore/f222b5ce-bef3-4c2c-bfda-f3c29bb77240?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4055409/tree/v2/index.html)

published in [Nature Medicine](/content/explore?query=Nature%20Medicine&refine=journal/index.html)

[Biology\\
\\
10 \| Feb \| 2026\\
\\
Analysis Code for "Bilingual language processing relies on shared semantic representations that are modulated by each language"\\
\\
Analysis code for "Bilingual language processing relies on shared semantic representations that are modulated by each language". Expected environment build time and analysis run time is around five minutes.\\
\\
Catherine Chen](/content/explore/8a08e874-0dc9-42db-8e92-91171ad229cf?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9997376/tree/v1/index.html)

[Physics\\
\\
29 \| May \| 2026\\
\\
Capsule for "Electron Ptychography Reveals Correlated Lattice Vibrations at Atomic Resolution"](/content/explore/97a3c8b0-c59d-4cf1-aeec-4cc9a8abc5b8?page=2&filter=all/index.html)

[This package contains data and scripts used in a paper "Electron Ptychography Reveals Correlated Lattice Vibrations at Atomic Resolution"](/content/explore/97a3c8b0-c59d-4cf1-aeec-4cc9a8abc5b8?page=2&filter=all/index.html) [https://doi.org/10.21203/rs.3.rs-7649135/v1](https://doi.org/10.21203/rs.3.rs-7649135/v1).
The package is divided in three folders:

1. phase retrival code pypty (written by Anton Gladyshev, AG SEM, Physics Department, Humboldt-Universität zu Berlin). The code can be found in the folder code/pypty
2. Raw 4DSTEM data used in the publication. The data can be found in the folder data/
3. .py files for the reconstructions as well as jupyter notebooks for the analysis of the data. The scripts can be found in the folder scripts/
4. Raw .npy results. The files for Silicon grain boundary can be found in the folder results\_silicon/, the results of the hBN reconstruction can be found in results\_hbn/

Anton Gladyshev et al.

[Open Capsule](/content/capsule/0448107/tree/v3/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Biology\\
\\
10 \| Feb \| 2026\\
\\
Enhancement of mediodorsal thalamus rescues aberrant belief dynamics in a mouse model with schizophrenia-associated mutation\\
\\
Beliefs drive decisions but require constant updating in order to match environmental dynamics. In schizophrenia this updating process is thought to be impaired leading to delusions, but the underlying neural substrates are unknown, in part, due to a lack of appropriate animal models and behavior readouts. Here, we address this challenge by taking two synergistic approaches. First, we generate a mouse model bearing patient-derived point mutation in Grin2a (Grin2aY700X+/-), a gene that confers high-risk for schizophrenia identified by recent large-scale exome sequencing. Second, we develop a computationally trackable foraging task, in which mice form and update belief-driven decision strategies in a dynamic environment. We found that Grin2aY700X+/- mice perform less optimally than their wild-type (WT) littermates, due to unstable cognitive states related to noisy representation of dynamic task values. Using functional ultrasound imaging, in vivo and ex vivo electrophysiological recording, we identified the mediodorsal (MD) thalamus as hypofunctional in Grin2aY700X+/- mice, and in vivo task recordings showed that MD neurons encoded dynamic task values and cognitive states in WT mice. Optogenetic inhibition of MD neurons in WT mice phenocopied Grin2aY700X+/- mice and enhancing MD activity rescued task deficits in Grin2aY700X+/- mice. Together, our study identifies the MD thalamus as a key node for schizophrenia-relevant cognitive dysfunction, and a potential target for future therapeutics.\\
\\
Tingting zhou](/content/explore/47914969-89c9-46b1-bbb2-f1cf9e01d8fe?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5554228/tree/v1/index.html)

published in [Nature Neuroscience](/content/explore?query=Nature%20Neuroscience&refine=journal/index.html)

[Social Sciences\\
\\
10 \| Feb \| 2026\\
\\
International Trade Reduces Emissions through Technology Transfer Led by Key Emitters\\
\\
This capsule includes MATLAB codes for data processing and visualization, Stata 18 cods for analysing the impact of agreements and technology on carbon emissions, and the processed panel data and derived patent transfer datasets.\\
\\
Jiaming Wang](/content/explore/d4545bd7-9ded-45ca-8c8c-dc9fd4b55b90?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0891615/tree/v1/index.html)

published in [Nature Climate Change](/content/explore?query=Nature%20Climate%20Change&refine=journal/index.html)

[Engineering\\
\\
9 \| Feb \| 2026\\
\\
Perch like a bird: bio-inspired optimal maneuvers and nonlinear control for Flapping-Wing Unmanned Aerial Vehicles\\
\\
The code simulates an ornithopter flying from initial conditions to a perching point autonomously, by the use of adaptive control throug an optimal perching trayectory.\\
\\
J. Á. Acosta & C. Ruiz-Páez](/content/explore/b2612283-93a3-441f-bc10-e36ec5d646c2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8864028/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/tcyb.2025.3650717) published in [IEEE Transactions on Cybernetics](/content/explore?query=IEEE%20Transactions%20on%20Cybernetics&refine=journal/index.html)

[Computer Science\\
\\
9 \| Feb \| 2026\\
\\
Zero-Shot Land Cover Classification of SAR-like Imagery using Vision-Language Models\\
\\
This capsule contains the reproducibility code for the IEEE GRSL submission "Zero-Shot Land Cover Classification of SAR-like Imagery using Vision-Language Models". It implements a comprehensive zero-shot classification framework using CLIP on the EuroSAT dataset. The code includes pipelines for: (1) converting optical imagery to SAR-like intensity, (2) evaluating 10+ prompt engineering strategies strategies (comparing spatial vs. domain-specific prompts), and (3) performing hierarchical classification (Coarse vs. Fine).\\
\\
Arivoli A, Ashik Sharon M & Mannadithya](/content/explore/72d8c92b-3814-411b-9173-2c1b4b7694cc?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2850219/tree/v1/index.html)

[Physics\\
\\
9 \| Feb \| 2026\\
\\
Transformer-Based Geometric Tomography for Complex-Energy Braiding Topology\\
\\
1. Test.py is used to generate Fig.3 in the main text, config.json, ft\_transformer.py, and utils.py are helper functions.\\
2. Visualization.py is used to generate the weight data for Fig.3 and Fig.5 in the main text, as well as\\
the supplementary Fig.3-Fig.10 in the Supplementary Material.\\
\\
Yang Yue et al.](/content/explore/17f5ddbb-9da5-4e17-8ca2-e0665669ee37?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6632247/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
7 \| Feb \| 2026\\
\\
Machine learning guided resolution of mechanical trade-off in polymer composites via stress adaptive interface\\
\\
This capsule provides a reproducible demonstration of a machine-learning–guided\\
framework for resolving mechanical trade-offs in polymer composites with\\
stress-adaptive interfacial architectures.\\
\\
The complete methodology integrates Gaussian process (GP) surrogate modeling,\\
Pareto set learning (PSL), and active learning (AL) to explore multi-objective\\
trade-offs among tensile strength, fracture toughness, and impact energy\\
dissipation. The framework is designed to efficiently navigate high-dimensional\\
composition spaces while minimizing experimental burden.\\
\\
In this capsule, we focus on the evaluation stage of the active learning\\
pipeline, using an initial experimental dataset (n = 50) as a demonstration.\\
The workflow includes GP model training, construction of an approximate Pareto\\
set, and quantitative assessment of convergence and stability using hypervolume\\
(HV) and stability (S) metrics. \\
\\
This capsule is intended to support transparency, reproducibility, and\\
methodological clarity for the accompanying manuscript.\\
\\
Hao Wang et al.](/content/explore/4deeefdb-46a4-4e4c-9c65-a0af7a392662?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1567350/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Bioinformatics\\
\\
7 \| Feb \| 2026\\
\\
Predicting and Interpreting Cell-type Specific Drug Responses in the Low data regime using Inductive Priors\\
\\
PrePR-CT is a graph-based deep learning method designed to predict transcriptional responses to chemical perturbations in single-cell data. This method utilizes Graph Attention Network (GAT) layers to encode cell-type graphs from batches of training samples. These encoded graphs are then integrated with control gene expression data and predefined perturbation embeddings. The combined data is processed through Multi-Layer Perceptrons (MLPs) to accurately predict gene expression responses. \\
\\
We have included the Kang dataset, which is essential for reproducing the results presented in the paper. Alongside the dataset, we provide a notebook example that demonstrates step-by-step instructions for processing the data and generating the figures featured in the publication.\\
\\
Reem Alsulami et al.](/content/explore/6ff8d8e9-336d-4db7-9106-40f71ab39af2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4423068/tree/v1/index.html)

published in [Nature Machine Intelligence](/content/explore?query=Nature%20Machine%20Intelligence&refine=journal/index.html)

[Engineering\\
\\
18 \| Mar \| 2026\\
\\
Investigating the merits of co-locating renewable energy with steel production to decarbonise the iron industry-Codes\\
\\
Scaling up renewable hydrogen for directly reduced iron (DRI-H) production in Electric Arc Furnaces (EAFs) may impose significant resource demands, particularly given the global iron ore trade. We modelled scenarios in which 1,468 Mt of iron ore – the total global export volume in 2021 – is processed using DRI-H and EAF technologies. Supplying renewable hydrogen via imported ammonia to existing steel-making facilities requires 35.42 EJ (range: 27.52-46.91 EJ) of solar energy, 135,882 km² (77,719-279,967 km²) of land use, and 5,768 GL (3269-8,545 GL) of water. By contrast, co-locating DRI-H facilities with solar PV farms – thereby avoiding the need to transport iron ore and other feedstocks globally – saves 50% in energy, 50% in land use, and 39% in water demand. This analysis illustrates the importance of aligning industrial decarbonisation with spatial planning and infrastructure development, and co-location may provide an opportunity for reducing the environmental impacts of global supply chains.\\
\\
Bishal Bharadaj et al.](/content/explore/127f5ec2-338a-4d16-9a23-dc4fb50b8e4b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5628208/tree/v3/index.html)

[Engineering\\
\\
6 \| Feb \| 2026\\
\\
PAD-Physics-Aware Differentiable Blind Source Separation: A Geometry-Guided Unsupervised Learning Framework\\
\\
This capsule reproduces the simulation results presented in the paper.\\
The code in PAD\_section5\_D.py corresponds to the simulation experiment setup in Section D of Chapter 5 (QPSK, FM, 16QAM, SNR=0 dB).\\
The code in PAD\_section5.py represents the simulation experiment configuration for Chapter 5, which randomly selects three signals from multiple options and assigns parameters randomly. This framework is compatible with CUDA execution.\\
\\
Mingdi Li](/content/explore/b6e1f5ab-90cc-4d06-99a6-cfa9c02ed627?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1977492/tree/v1/index.html)

[Engineering\\
\\
6 \| Feb \| 2026\\
\\
Small-Signal Modeling and Analysis of a Grid-Forming PEM Hydrogen Electrolyzer\\
\\
This code generates the small-signal model of a grid-forming PEM hydrogen electrolyzer system. It utilizes the parameters of the PEM electrolyzer, the DC-DC buck converter, the DC-AC inverter, and the operating point around which the small-signal model is linearized. The output of this script is a linearized state-space representation suitable for analyzing the dynamic behavior of the PEMEL unit. This model is particularly useful for studying integration scenarios involving other loads and energy resources within power systems.\\
\\
Basil Hamad, Yasser Abdel-Rady I. Mohamed, Ahmed Al-Durra & Ehab El-Saadany](/content/explore/fe73c176-f330-495c-a25d-831d8374de84?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1603263/tree/v1/index.html)

[Bioinformatics\\
\\
6 \| Feb \| 2026\\
\\
De novo Design of Functional Nucleic Acids of Aptamers\\
\\
This is a framework to use nucleic acids language model to generate functional nucleic acids.\\
\\
Zhiming Zhang et al.](/content/explore/0e35045f-ac0a-4372-8905-016e072b9553?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1677781/tree/v1/index.html)

published in [Nature Computational Science](/content/explore?query=Nature%20Computational%20Science&refine=journal/index.html)

[Biology\\
\\
6 \| Feb \| 2026\\
\\
Cortex-wide characterization of decision-making neural dynamics during spatial navigation\\
\\
This capsule includes the main data analysis codes used in "Cortex-wide characterization of decision-making neural dynamics during spatial navigation". Sample data is provided for running the k-means clustering code used in this manuscript to identify cortical activation states (brain states). A refined sample dataset is also provided to test data visualization and motif analysis.\\
\\
Samuel Haley](/content/explore/cd462a14-cfb3-4445-b8e2-5a8238f760bf?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6036145/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
5 \| Feb \| 2026\\
\\
Exact Outage Probability and Ergodic Capacity of NOMA\\
\\
This capsule provides reproducible code for the exact performance analysis of a two-user downlink non-orthogonal multiple access (NOMA) system with successive interference cancellation (SIC). The implementation follows an exact post-SIC modeling framework in which the disturbance observed at the near user is modelled as a truncated-Gaussian noise process conditioned on SIC success and failure events. The capsule computes the near-user outage probability by explicitly decomposing performance over SIC outcomes and averaging over Rayleigh fading. It also evaluates the near-user ergodic capacity using the same post-SIC SINR characterization. Monte Carlo simulations are used to validate the analytical and numerical results and to compare them with conventional legacy Gaussian and imperfect-SIC models. All key system parameters, such as power allocation factors, target rates, SNR range, and channel statistics, are user-configurable so that the numerical results reported in the manuscript can be reproduced and extended.\\
\\
Alok Kumar Shukla, Arafat Al-Dweik & Sami Muhaidat,](/content/explore/3c17aa73-83f5-49c7-b184-24954193135e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0972286/tree/v1/index.html)

[Bioinformatics\\
\\
5 \| Feb \| 2026\\
\\
Myofibroblasts induce neuroplasticity to promote pancreatic inflammation and cancer progression \[Functional Enrichment on PCSs Differentially Expressed Genes from Mucciolo G. et al 2024\]](/content/explore/98f688c8-a4e4-48f5-aab7-3391be3f06aa?page=2&filter=all/index.html)

[This capsule contains the functional enrichment analysis conducted on the differentially expressed genes obtained from the \\
bulk RNA-seq PSCs as published in](/content/explore/98f688c8-a4e4-48f5-aab7-3391be3f06aa?page=2&filter=all/index.html) [Mucciolo et al. 2024](https://doi.org/10.1016/j.ccell.2023.12.002).
The capsule also includes the code to generate an enrichment map with the
selected terms of interest related to axon obtained during the functional enrichment step.

More specifically, the data used in this analysis are from [Mucciolo et al. 2024](https://doi.org/10.1016/j.ccell.2023.12.002):

- _Table S1._ RNA-sequencing analysis of control-treated or TGF-beta-treated PSCs - differential expression analysis (related to Figure 1 in Mucciolo et al. 2024).
- _Table S2._ RNA-sequencing analysis of control-treated or CM-treated PSCs with or without ERBBi - differential expression analysis (related to Figures 2 and 3 in Mucciolo et al. 2024).

Astrid Deschênes, Jérémy Nigri & David A. Tuveson

[Open Capsule](/content/capsule/5037799/tree/v1/index.html)

[Mathematics\\
\\
5 \| Feb \| 2026\\
\\
Physics-informed AI for cross-regional epidemic spillover: a two-region framework that guides non-outbreak interventions\\
\\
This repository contains the source code, data, and instructions for replicating the results of our manuscript "Physics-informed AI for cross-regional epidemic spillover: a two-region framework that guides non-outbreak interventions". The TDINN framework integrates deep neural networks with mechanistic epidemic models to infer time-varying parameters such as contact rates and bidirectional spillover rates across regions.\\
\\
Hao Wang et al.](/content/explore/33e8e6e1-d592-401c-bcc0-028e53807178?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6742575/tree/v1/index.html)

[Physics\\
\\
5 \| Feb \| 2026\\
\\
A nutrient bottleneck controls antibiotic efficacy in structured bacterial populations](/content/explore/63ec11bb-168c-45fc-9e02-7962e9320a40?page=2&filter=all/index.html)

[The included code includes MATLAB function lines.m and script runsim\_codeocean.m for simulations reported in the manuscript “A nutrient bottleneck controls antibiotic efficacy in structured bacterial populations” by Anna M. Hancock, Arabella S. Dill-Macky, Jenna A. Moore, Catherine Day, Mohamed S. Donia, and Sujit S. Datta, submitted for consideration as an article in Nature Communications.\\
\\
Antibiotic resistance is a growing global health threat. Although antibiotic activity is well studied at the single cell level, many infections are caused by spatially structured multicellular populations where consumption of scarce nutrients establishes strong spatial variations in their abundance. These nutrient variations have long been hypothesized to help bacterial populations tolerate antibiotics, since single-cell studies link antibiotic tolerance to metabolic activity, and thus, local nutrient availability. Here, we test this hypothesis by visualizing cell death in structured \\textit{Escherichia coli} populations exposed to nutrient (glucose) and antibiotic (fosfomycin). We find that nutrient availability acts as a bottleneck to antibiotic killing, causing death to propagate through the population as a traveling front. By integrating our measurements with biophysical theory and simulations (sample code included here), we establish quantitative principles that explain how collective nutrient consumption can limit the progression of this \`\`death front,'' protecting a population from a nominally deadly antibiotic dose. While increasing nutrient supply can overcome this bottleneck, in some cases, excess nutrient unexpectedly \\emph{promotes} the regrowth of resistant cells. Altogether, this work provides a key step toward predicting and controlling antibiotic treatment of spatially structured bacterial populations, yielding biophysical insights into collective behavior and guiding strategies for effective antibiotic stewardship.\\
\\
The parameters in runsim\_codeocean.m, the code run in this demo, recreate data reported in Fig. 3E-G. By changing the input parameters within the script, this same code, which calls the function lines.m to evaluate the model reported in the manuscript, can generate simulation results reported elsewhere in the manuscript.](/content/explore/63ec11bb-168c-45fc-9e02-7962e9320a40?page=2&filter=all/index.html)

[All code was demo’ed on MATLAB 2024a, which can be installed following directions found here:](/content/explore/63ec11bb-168c-45fc-9e02-7962e9320a40?page=2&filter=all/index.html) [https://www.mathworks.com/help/install/index.html](https://www.mathworks.com/help/install/index.html)

Please contact Anna Hancock ( [annahancock@princeton.edu](mailto:annahancock@princeton.edu)) with questions.

Anna M. Hancock et al.

[Open Capsule](/content/capsule/9055864/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Bioinformatics\\
\\
4 \| Feb \| 2026\\
\\
Subtype-DIVIDE: A Self-supervised Diffusion Contrastive Framework for Multi-omic Cancer Subtyping\\
\\
Subtype-DIVIDE is a self-supervised diffusion contrastive learning framework for multi-omic cancer subtyping. The capsule provides code to preprocess multi-omic data, train the Subtype-DIVIDE model, and evaluate discovered cancer subtypes using downstream clustering and survival/clinical analyses.\\
This compute capsule is intended to support reproducibility of the experiments described in the associated manuscript. It includes the main training/inference script and helper utilities, along with instructions to run the pipeline on supported datasets.\\
\\
Liting Zhang, Zefeng Li & Xian Mallory](/content/explore/633438a5-c5e9-4d3f-9aef-5d45b1f1bc96?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2285848/tree/v1/index.html)

[Computer Science\\
\\
4 \| Feb \| 2026\\
\\
Sustainable, QoS, and Cost-Aware Placement of Microservices on the Continuum: A Use Case on the Internet of Medical Things\\
\\
Domains such as healthcare, which are intensive in contrast to user-grade domains, are increasingly interested to develop Internet of Things (IoT) applications to automate their critical processes, leading to the Internet of Medical Things (IoMT). IoMT applications leverage the Computing Continuum infrastructure. However, for intensive IoT and IoMT applications to be feasible, their strict Quality of Service (QoS) requirements must be met, and the economic cost of their deployment must be low to ensure their business viability. Furthermore, sustainability and the reduction of the carbon footprint are currently a priority due to industrial and governmental initiatives and requirements, such as Green IoT or the Sustainable Development Goals. The efforts in sustainability are aimed not only at achieving energy-efficient applications, but carbon-aware ones, aligning carbon footprint reduction with high QoS and low economic cost. While all three objectives can be achieved by strategically placing the microservices of these IoT applications, navigating the trade-offs across the three is a complex issue, calling for automated solutions that provide IoT application developers with a manageable number of Pareto-optimal microservice placements. This work presents the Many-Objective Genetic Algorithm for Microservice Placement (MOGAMP), which leverages evolutionary computing to assist IoT application developers in navigating the QoS, cost, and sustainability trade-off in microservice placement. In an evaluation with an IoMT use case, MOGAMP is shown to be scalable, up to 459.82 times faster and with a memory footprint of up to 0.37% compared to alternatives, enabling IoT application developers to explore wide, yet manageable, Pareto fronts.\\
\\
Acknowledgment\\
This work has been partially funded by the European Union under the MSCA project RENOS (contract number 101205037). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Executive Agency. Neither the European Union nor the granting authority can be held responsible for them. This work has also been partially funded by the projects TED2021-130913B-I00, PDC2022-133465-I00, PID2022-140907OB-I00, PRTR-C17.I1 and RED2022-134337-T funded by MICIU/AEI/10.13039/501100011033, the European Union "Next Generation EU/PRTR" and ERDF "A way of making Europe" as well as by the Cap4IE project (0786\_CAP4ie\_4\_P) funded by the Interreg V-A España-Portugal (POCTEP) 2014-2020 program, by the Regional Ministry of Economy, Science and Digital Agenda of the Regional Government of Extremadura (GR21133) and the European Regional Development Fund.\\
\\
Alejandro Moya et al.](/content/explore/a682f3d9-0aa5-4e56-b428-3d57d9d4663b?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7928794/tree/v1/index.html)

[Biology\\
\\
25 \| Mar \| 2026\\
\\
Biodiversity resilience in a tropical rainforest\\
\\
This code calculates recovery times, resistance and return rates of biodiversity from data sampled in Reserva Canandé, Ecuador belonging to the manuscript "Biodiversity resilience in a tropical rainforest". Furthermore, all figures that are part of the corresponding papers are created with this code.\\
\\
In this repository, we provide all raw data used for this analysis, as well as calculated values and scripts that were used during the analysis.\\
\\
We provide raw data of observed species per plot in the Folder "Data/Raw\_data". Data is either in long format or wide format, depending on taxon.\\
\\
The script "Calculate\_metrics.R" produces a table for each taxon containing for each plot, the abundance, Shannon diversity (alpha\_q1) and Bray-Curtis distance. In this file we also additionally report the treatment of the plot (active, regeneration or old-growth forest), time since start of recovery (RegTime) and the legacy (cacao, pasture or old-growth). \\
\\
The script "Calculate\_metrics\_standardized.R" produces a table for each taxon containing for each plot the abundance, species richness (alpha\_q0), Shannon diversity (alpha\_q1), Simpson diversity (alpha\_q2), standardized species richness (alpha\_q0\_std), standardized Shannon diversity (alpha\_q1\_std), standardized Simpson diversity (alpha\_q2\_std), Jacccard-index (beta\_q0), Horn-index (beta\_q1), Morisita-Horn-index (beta\_q2), standardized Jaccard-index (beta\_q0\_std), standardized Horn-index (beta\_q1\_std), standardized Morisita-Horn-index (beta\_q2\_std). In this file we also additionally report the treatment of the plot (active, regeneration or old-growth forest), time since start of recovery (RegTime) and the legacy (cacao, pasture or old-growth). \\
\\
The Python-Script "Calculate\_recoverytimes.py" was used to calculate recovery times, resistance and return rates for the metrics reported in files created with the scripts "Calculate\_metrics.R". The script fits the negative exponential model given by Eq. 5 to the input dataset and calculates recovery times according to Eq. 6 and resistance according to Eq. 4. The parameter lambda is optimized and used as the return rate. The results are reported in the file "Supplementary\_Table\_1.csv". We used the Python-Script "Calculate\_standardized\_recoverytimes.py" to calculate recovery times, resistance and return rates for the standardized metrics produced by the script "Calculate\_metrics\_standardized.R". The results are reported in the file "Supplementary\_Table\_3.csv". The files "Supplementary\_Table\_1.csv" and "Supplementary\_Table\_3.csv" contain the columns R2, which contains the r-squared value of the fitted function, "RSE", which contains the Residual Standard Error and "Active\_equals\_OG", which indicates whether the mean value of the respective taxon and index of all the active agriculture (or only cacao or pasture) plots are similar to the mean of the old-growth forest plots within the standard deviations.\\
\\
Random-Forest results for our data were created with the "Random\_forest\_analysis\_legacy.R"-script.\\
\\
Statistical tests to test for the influence of life-history-strategies, trophic level and mobility were done with the "Explain\_recoverytimes\_legacy.R"-script, which uses Supplementary Tables 1 and 4 as input.\\
\\
Statistical tests to test whether standardized or non-standardized metrics recover quicker were done with the "Standardized\_vs\_non-standardized\_legacy.R"-script.\\
\\
For each study identified in the literature analysis the alpha-diversity Hill-number (q=0) and the beta-diversity Hill-number (q=0) were calculated for each plot using the script "Calculate\_metrics\_literature.R". The results for each study are stored in the folder "Literature/Data/Taxon" under the respective name of the reference.\\
\\
The code used to calculate these values is stored in "Calculate\_recoverytimes\_literature.ipynb".\\
\\
Random-Forest results for the literature data was created with the "Random\_forest\_analysis\_literature.R"-script.\\
\\
Figure scripts are named as the figure they produce, apart from Figures 3 and Extended Data Figs. 1 and 2, which were created within the "Calculate\_recoverytimes.py"-script.\\
\\
The values and confidence intervals for resistance in Supplementary Tables 1, 3 and 5 need to be multiplied with 100 for a percentage.\\
\\
Nico Blüthgen et al.](/content/explore/e68245b5-78c5-460f-aa17-1db3a10c3b0d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6264306/tree/v5/index.html)

published in [Nature](/content/explore?query=Nature&refine=journal/index.html)

[Computer Science\\
\\
4 \| Feb \| 2026\\
\\
Sustainable Federated Learning in 6G Networks: The Case for Cybersecurity\\
\\
6G networks are envisaged to be AI-native and employ distributed AI at various layers, which would enable **Federated Learning (FL)** as an integral technology. Despite the significant benefits, the **carbon footprint** of FL in such complex systems warrants deeper inspection and optimization as part of sustainability concerns. Moreover, as next-generation networks face emerging threat vectors, **cybersecurity** plays a crucial role in defending against such attacks. This study integrates the concepts of FL, sustainability, and cybersecurity in the context of future networks, and presents a use-case of intrusion detection on a **Virtual Network Function (VNF)** dataset.\\
\\
In the scope of the study, this capsule provides a codebase for conducting _carbon-aware_ FL experiments using the **Flower** framework, where each client employs **CodeCarbon** to measure and track their carbon footprint during the local training phase of an FL pipeline. By using FL, a global intrusion detection model is trained and evaluated on VNF traffic. Experiments are conducted across different _carbon-aware client selection_ mechanisms, and the results are compared based on the intrusion detection model performance and the carbon footprint of the training process.\\
\\
Gökcan Cantali, Wissem Soussi, Gürkan Gür & Burkhard Stiller](/content/explore/8710b98b-5ba5-48ec-9a27-59903c606e7d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8012801/tree/v1/index.html)

[Mathematics\\
\\
4 \| Feb \| 2026\\
\\
MooAFEM and octAFEM3D: Computational study of the smoothed adaptive finite element method (S-AFEM) for second-order elliptic PDEs\\
\\
We investigate the smoothed adaptive finite element method (S-AFEM) proposed in \[Mulita, Giani, Heltai: SIAM J. Sci. Comput. 43, 2021\]. S-AFEM modifies the classical adaptive finite element method (AFEM) by performing accurate discrete solves only on periodically determined mesh levels, while the intermediate levels employ a fixed number of cheap smoothing iterations. This strategy generates adapted meshes comparable to those of AFEM at substantially lower computational cost. We consider standard smoothers such as Richardson, Gauss-Seidel, conjugate gradient, and multigrid schemes. The algorithm ensures unconditional full R-linear convergence and, for sufficiently small adaptivity parameters, optimal convergence rates with respect to the overall computational cost for general second-order linear elliptic PDEs. The 2D experiments were conducted using the object-oriented Matlab software package MooAFEM. For the experiments in 3D, we extended the octAFEM3D software package. The numerical experiments reproduce the results reported in the associated publication. They validate the convergence results, analyze runtime performance, and underline the potential of S-AFEM for speed-up in AFEM computations.\\
\\
Philipp Bringmann, Christoph Lietz & Dirk Praetorius](/content/explore/71dab42c-b131-4a6e-9214-c526691eb3fd?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0412224/tree/v1/index.html)

[Earth Sciences\\
\\
4 \| Feb \| 2026\\
\\
Region-specific and nutritionally adequate dietary transitions can bolster sustainability and socioeconomic benefits\\
\\
This repository contains all materials necessary to reproduce the figures from the manuscript _Environmental and Societal Implications of Transitioning to Sustainable Diets_. It includes the analysis scripts, plotting code, and all required datasets. To generate the figures, simply run the script `run`, which will launch `paper_analysis.R` and `paper_methodology.R`.\\
\\
Clàudia Rodés-Bachs et al.](/content/explore/1f833012-856f-477d-a6b2-20a872b3197e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8530334/tree/v1/index.html)

published in [Nature Food](/content/explore?query=Nature%20Food&refine=journal/index.html)

[Bioinformatics\\
\\
3 \| Feb \| 2026\\
\\
Aberrant low expression of LDHB in fibroblasts promotes breast cancer metastasis via lactate-mediated inflammatory reprogramming\\
\\
This comprehensive script repository is designed to reproduce the code, data processing, and visualization for multiple figures of the manuscript, performing comprehensive metabolic pathway analysis using Gene Set Variation Analysis (GSVA) with single-sample GSEA methodology on TCGA-BRCA RNA-seq data to generate pathway activity scores that form the basis for the metabolic heatmaps and pathway enrichment patterns displayed in Figures 1A and 1C, while also reproducing the single-cell RNA sequencing analyses and visualizations for Figures 1I, 1J, and 1K, including UMAP dimensionality reduction for cell type identification, feature expression plotting of metabolic genes, and cell type-specific expression analysis in fibroblast populations to demonstrate cellular heterogeneity and metabolic gene expression patterns in the tumor microenvironment, with all code generating publication-quality figures directly used in the manuscript.\\
\\
zhihong luo](/content/explore/dc7c7071-2fe1-4df0-bfd6-d511e878a5a6?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7411670/tree/v1/index.html)

[Mathematics\\
\\
3 \| Feb \| 2026\\
\\
Randomized Projection Operators onto piecewise Polynomial Spaces\\
\\
This Jupyter notebook accompanies the paper “Randomized Projection Operators onto piecewise Polynomial Spaces.”\\
It contains all scripts required to reproduce the figures and tables in the manuscript. In particular, the code maps given right-hand sides onto piecewise polynomials via random point-evaluations. Then it solves the Poisson model problem with the approximated piece-wise polynomial right-hand side and compares it to deterministic evaluations.\\
\\
Johannes Storn](/content/explore/f66adc7b-1b18-4070-8dc1-16f4e3ca96f2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0753226/tree/v1/index.html)

[Bioinformatics\\
\\
3 \| Feb \| 2026\\
\\
AI-driven large-scale electron microscopy enables whole-tissue subcellular digitization\\
\\
Abstract: The distribution and interactions of organelles critically mediate cellular physiology and pathology. Large-scale electron microscopy enables visualization of subcellular networks at the tissue level with nanometer resolution, but robust and efficient computational tools are lacking. Here, we present DeepOrganelle, a deep learning tool for high-throughput, cell-resolved mapping and digitization of organelle distribution and interactions in large-scale 2D/3D electron microscopy. When applied to spermatogenesis across 22 differentiation status of the germ cells, DeepOrganelle uncovered previously unrecognized, stage-dependent dynamics of membrane contact sites within one subphase during meiosis. It also revealed coordinated organelle redistribution in Sertoli cells towards the blood–testis barrier, digitizing the remodeling dynamics of the tissue. This study establishes DeepOrganelle as a powerful framework for capturing subcellular dynamics at the whole-tissue level.\\
\\
Li Xiao et al.](/content/explore/d696211d-da7d-4249-8eb0-81fca2d5f923?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2577186/tree/v1/index.html)

[Engineering\\
\\
3 \| Feb \| 2026\\
\\
Multi-Rate NOMA with Arbitrary Number of users and symbol Rates\\
\\
This capsule provides the MATLAB code used to generate the theoretical and simulation results presented in the paper \\
“Multi-Rate NOMA with Arbitrary Number of Users and Symbol Rates.” Running the main script reproduces the performance curves shown in the paper.\\
\\
Zainab Khader, Arafat Al-Dweik & Emad Alsusa](/content/explore/045d5273-6b14-4530-842b-bf4956011e6a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8660670/tree/v1/index.html)

[Chemistry\\
\\
1 \| Jun \| 2026\\
\\
Machine-learning enhanced simulations predict graphene is microscopically hydrophobic and not wetting transparent](/content/explore/474d0963-5cbe-4037-890d-218e0acab15d?page=2&filter=all/index.html)

[This script is primarily used to analyze the water orientation in our graphene/water systems. To run the script, using the command "julia angle\_orientation.jl >> angle.txt 2>&1 &"](/content/explore/474d0963-5cbe-4037-890d-218e0acab15d?page=2&filter=all/index.html)

[If you use the script in your work, please cite the associated publication:](/content/explore/474d0963-5cbe-4037-890d-218e0acab15d?page=2&filter=all/index.html) [https://doi.org/10.1038/s41467-026-71053-3](https://doi.org/10.1038/s41467-026-71053-3)

Dianwei Hou

[Open Capsule](/content/capsule/6138159/tree/v2/index.html)

[Associated article](https://doi.org/10.1038/s41467-026-71053-3) published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
2 \| Feb \| 2026\\
\\
Approximate Strategy for Stochastic Mean-Field Incentive Stackelberg Games with Delay\\
\\
We provide a numerical program and an appendix demonstrating the performance of the approximate strategies for the incentive Stackelberg game in time-delayed stochastic mean-field systems with a large population.\\
\\
Hiromu Ozai](/content/explore/a52014d5-380e-45e0-86ec-f8202aa91146?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2619623/tree/v1/index.html)

[Engineering\\
\\
2 \| Feb \| 2026\\
\\
Decentralized Stochastic Constrained Optimization via Prox- Linearization: Symbolic Simplification Codes\\
\\
The uploaded MATLAB scripts are used for symbolic and algebraic simplification of analytical expressions arising in the theoretical analysis of algorithms proposed in paper Decentralized Stochastic Constrained Optimization via Prox-Linearization.\\
Note: These scripts are intended only for proof verification and\\
simplification and do not implement or simulate the proposed\\
algorithms.\\
Files are\\
• dsmpl\_2.m: Symbolic simplification of convergence and error bounds for the D-SMPL algorithm.\\
• dscampl\_2.m: Symbolic simplification of convergence bounds for the D-SCAMPL algorithm.\\
Usage\\
The scripts are used to: - encode analytical bounds symbolically, - substitute theoretical\\
parameter selections, - simplify expressions, (Lemma 5, Lemma 11)and\\
\\
- verify asymptotic convergence rates (Theorem 1 and Theorem 2).\\
Requirements\\
• MATLAB\\
• Symbolic Math Toolbox\\
\\
Basil M. Idrees, Shivangi Dubey, Lavish Arora & Ketan Rajawat](/content/explore/b931985e-9707-4bb3-ac3b-3617896f0567?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2189258/tree/v1/index.html)

[Earth Sciences\\
\\
2 \| Feb \| 2026\\
\\
Physical limits of sea-level rise adaptation in global river deltas: code instruction\\
\\
Abstract: Sea-level rise threatens deltas worldwide, requiring adaptation to flood risks. Delta adaptation is typically presented as a choice between five strategies: advance, protect-closed, protect-open, accommodate, and retreat. However, a full assessment of the physical feasibility of these strategies across deltas remains limited. We present a first-order assessment of the physical solution space for adaptation to sea-level rise for nearly 800 deltas globally. We find that current technologies, resources, and space provide at least one physically feasible delta-wide strategy for every delta to adapt by 2100. This number may increase in the future through technical innovation or collaboration between deltas. The type and number of physically feasible strategies are mostly determined by delta’s physical characteristics, whereby, large, urbanized, or frequently flooded deltas have fewer options than small, rural, or infrequently flooded deltas. Our analysis highlights the risk of resource limitations as global deltas will need to adapt simultaneously to future flood risks.\\
\\
This capsule contains the code and datasets used to map the physical solution space of global deltas by 2100 under a number of climate scenarios. The workflow combines data preprocessing, modelling and visualizations. The model determines the physical feasibility of adaptation by assessing the physical characteristics of each delta and applying three thresholds to assess whether measures can be implemented. The visualization scripts include the physical solution space figures for individual and global deltas, and the material requirements for each strategy.\\
\\
Kiara Lasch, Jaap Nienhuis, Gundula Winter & Marjolijn Haasnoot](/content/explore/06e6320f-9b17-4f3f-b5e6-05f1364720b1?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8754340/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Physics\\
\\
2 \| Feb \| 2026\\
\\
Code for Cosmic CO and \[CII\] backgrounds chart star-forming gas over 12 Gyr](/content/explore/b2560973-e531-42db-b337-e14104b72506?page=2&filter=all/index.html)

[This package contains the code and data accompanying the paper “Cosmic CO and \[CII\] backgrounds chart star-forming gas over 12 Gyr” by Y.-K. Chiang. The paper presents the first intensity mapping detections of the cosmic mean amplitudes of the CO rotational ladder and \[CII\] 158 µm line, using the tomographic cross-correlation–based data vector and covariance matrix from Chiang, Makiya, & Menard 2025 (](/content/explore/b2560973-e531-42db-b337-e14104b72506?page=2&filter=all/index.html) [https://doi.org/10.3847/1538-4357/adfb6a](https://doi.org/10.3847/1538-4357/adfb6a)) released at [https://doi.org/10.5281/zenodo.16486649](https://doi.org/10.5281/zenodo.16486649).

To run MCMC chains for the full model fitting and the per-redshift refits, execute codes 01, 02, and 03. The chain lengths are set shorter than those used in the paper to allow reasonable run times. Pre-computed full chains are provided under the data directory.

To reproduce the key results and plots on line luminosity densities, the molecular gas density parameter Ω\_H2​​, and the mean line brightness temperature T\_b, execute codes 04 and 05.

Summary of included data:
(1) Tomographic CIB data vector and covariance matrix from Chiang et al. (2025; [https://doi.org/10.3847/1538-4357/adfb6a](https://doi.org/10.3847/1538-4357/adfb6a)), released at [https://doi.org/10.5281/zenodo.16486649](https://doi.org/10.5281/zenodo.16486649)

- CIB\_emissivity\_b.ecsv
- CIB\_emissivity\_b\_covariance\_matrix.txt

(2) CIB monopoles and errors from FIRAS + Planck measured in Odegard et al. (2019; [https://doi.org/10.3847/1538-4357/ab14e8](https://doi.org/10.3847/1538-4357/ab14e8)).

- CIB\_monopole\_Odegard\_19.fits

(3) Pre-computed MCMC posterior chain from this work for the full 21-parameter CIB + CO + \[CII\] model constrained by the data above.

- MCMC\_parameter\_posterior\_full.p

(4) Pre-computed MCMC posterior chains from this work for the per-redshift CO and \[CII\] refits constrained by the data above.

- MCMC\_parameter\_posterior\_per\_redshift\_CO.p
- MCMC\_parameter\_posterior\_per\_redshift\_CII.p

(5) Pre-computed mean line monopole brightness temperature derived from the full MCMC chain, released at [https://doi.org/10.5281/zenodo.16750940](https://doi.org/10.5281/zenodo.16750940)

- Tb\_cosmic\_CO\_CII.ecsv

Yi-Kuan Chiang

[Open Capsule](/content/capsule/6264752/tree/v1/index.html)

published in [Nature Astronomy](/content/explore?query=Nature%20Astronomy&refine=journal/index.html)

[Chemistry\\
\\
30 \| Jan \| 2026\\
\\
Structural defects in amyloid-β fibrils drive secondary nucleation\\
\\
Uses standard Python packages to fit a simple quadratic ligand binding model to FCS data on Brichos binding to Ab40 fibrils.\\
\\
Jing Hu](/content/explore/995ddf4e-bed5-438b-b91b-4bcabf8ac37e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9535526/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Economics\\
\\
2 \| Feb \| 2026\\
\\
Robust Automated Ensemble Selection for Time Series Forecasting: The UASEP Framework\\
\\
UASEP is introduced as a fully automated framework for unsupervised ensemble forecasting.\\
This estimator is based on the median of all solutions proposed by the alPCA algorithm, so that the end user has a single reliable and robust solution.\\
It is designed to obtain predictions for each of the time series included in the M3 competition. All you have to do is change the number of the time series, s, at the beginning of the code.\\
\\
Carlos García-Aroca, Mª Asunción Martínez Mayoral, Javier Morales Socuéllamos & José Vicente Segura Heras](/content/explore/32ea60f4-8451-41d2-9f4b-7dc8b342b6b7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5449379/tree/v1/index.html)

[Computer Science\\
\\
20 \| May \| 2026\\
\\
Code: Learning Incentive Structures for Cooperative Resilience in Multi-Agent Systems under Social Dilemmas\\
\\
This repository accompanies the paper "Learning Reward Functions for Cooperative Resilience in Multi-Agent Systems", which investigates how reward function design impacts cooperative resilience in Multi-Agent Reinforcement Learning (MARL).\\
\\
In dynamic and failure-prone environments, agents must not only optimize individual objectives but also ensure the collective system remains functional under disruptions. We define cooperative resilience as the ability of agents to anticipate, resist, recover, and adapt in the presence of external shocks. This repository provides tools and experiments to study and improve this emergent property through IRL-guided reward learning.\\
\\
We introduce a novel reward learning framework that learns reward functions from ranked trajectories—evaluated via a cooperative resilience score. Agents are then trained in social dilemma environments using:\\
\\
(i) Traditional individual reward functions\\
(ii) Inferred rewards aligned with cooperative resilience\\
(iii) Hybrid rewards combining both\\
The reward inference is performed using two preference-based IRL algorithms across three types of parameterizations:\\
\\
Handcrafted features\\
Linear reward models\\
Neural networks\\
Our results show that resilience-guided rewards lead to improved robustness and coordination, helping agents avoid catastrophic outcomes (e.g., resource depletion), without sacrificing individual performance. We further extend the experiments to larger 16×16 environments with 4 agents and multiple resource clusters, and evaluate agents under three disruption protocols (resource removal, regeneration slowdown, and agent perturbation).\\
\\
Manuela Chacon-Chamorro, Luis Felipe Giraldo & Nicanor Quijano](/content/explore/9e08b26c-f0ca-433a-bd18-cf6992a3a531?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4898801/tree/v1/index.html)

[Engineering\\
\\
2 \| Feb \| 2026\\
\\
Approximate Projections onto the Positive Semidefinite Cone Using Randomization](/content/explore/dcabedad-112b-4588-9b67-0066c4f90d24?page=2&filter=all/index.html)

[This repository contains MATLAB code for benchmarking randomized algorithms for projecting real symmetric matrices onto the Positive Semidefinite (PSD) cone using Randomized Numerical Linear Algebra (RNLA).](/content/explore/dcabedad-112b-4588-9b67-0066c4f90d24?page=2&filter=all/index.html)

[The implementation accompanies the paper:\\
Approximate Projections onto the Positive\\
Semidefinite Cone Using Randomization](/content/explore/dcabedad-112b-4588-9b67-0066c4f90d24?page=2&filter=all/index.html) [https://arxiv.org/pdf/2410.19208](https://arxiv.org/pdf/2410.19208)

Morgan Jones & James Anderson

[Open Capsule](/content/capsule/6132215/tree/v1/index.html)

[Computer Science\\
\\
30 \| Jan \| 2026\\
\\
NNV3: Expanding Neural Network Verification to New Architectures and Domains](/content/explore/86ee2287-42fb-4f53-9694-26a2d649e5c2?page=2&filter=all/index.html)

[This is the repeatability evaluation for NNV3: The Neural Network Verification Tool.](/content/explore/86ee2287-42fb-4f53-9694-26a2d649e5c2?page=2&filter=all/index.html)

[Samuel Sasaki, Anne M. Tumlin, Diego Manzanas Lopez, Muhammad\\
Usama Zubair, Navid Hashemi, Hongchao Zhang, Ben Wooding, Waseem\\
Abbas, and Taylor T. Johnson (](/content/explore/86ee2287-42fb-4f53-9694-26a2d649e5c2?page=2&filter=all/index.html) [http://www.taylortjohnson.com/](http://www.taylortjohnson.com/))

VeriVITAL - The Verification and Validation for Intelligent and Trustworthy Autonomy Laboratory ( [http://www.VeriVITAL.com](http://www.verivital.com/))
Institute for Software Integrated Systems (ISIS): [http://isis.vanderbilt.edu](http://isis.vanderbilt.edu/)
Electrical Engineering and Computer Science (EECS): [https://engineering.vanderbilt.edu/eecs/](https://engineering.vanderbilt.edu/eecs/)
Vanderbilt University: [https://www.vanderbilt.edu/](https://www.vanderbilt.edu/)

Source code and latest release:
[https://github.com/verivital/nnv](https://github.com/verivital/nnv)

Ben Wooding

[Open Capsule](/content/capsule/6810863/tree/v1/index.html)

[Computer Science\\
\\
29 \| Jan \| 2026\\
\\
Deep Learning-Based Modeling of Solar Drying Kinetics of Charal (Chirostoma spp.) for Sustainable Fish Protein Preservation\\
\\
This study investigates the low-temperature solar drying kinetics \\
of whole Charal under forced convection, natural convection, and open-air conditions. Experimental drying \\
curves were modeled using classical thin-layer formulations, and the most accurate models were subsequently \\
integrated into a data-driven framework.\\
\\
César A. García García](/content/explore/3c7aeb71-27b7-4554-83c8-3d56cb92fa10?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0698508/tree/v1/index.html)

[Biology\\
\\
29 \| Jan \| 2026\\
\\
Engineering non-exponential proliferation in Escherichia coli using functionalized protein aggregates\\
\\
Individual based model for the manuscript Engineering non-exponential proliferation in Escherichia coli using functionalized protein aggregates\\
\\
Ronald Van Eyken et al.](/content/explore/5454fa10-9bfa-4f72-b755-0bf2a76b6ab7?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1884928/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Earth Sciences\\
\\
2 \| Mar \| 2026\\
\\
Tropical cyclone rainfall extends inland\\
\\
This capsule is for you who want to reproduce the plots in the article "Tropical cyclone rainfall extends inland in a warming and urbanising world". All required packages are installed, and you can click run to plot all results.\\
\\
E Deng et al.](/content/explore/747ff3c8-2378-4e24-ac03-4a968bdedb52?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0568189/tree/v2/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Bioinformatics\\
\\
29 \| Jan \| 2026\\
\\
Conditional Diffusion with Locality Aware Modal Alignment for Generating Diverse Protein Conformational Ensembles\\
\\
This package contains the data and the code for the MacDiff algorithm (Modal-aligned conditional Diffusion) to predict the protein conformational ensembles.\\
\\
Baoli wang et al.](/content/explore/1b726e8f-e9e3-4143-a587-dc70d2aa389d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5363285/tree/v1/index.html)

published in [Nature Machine Intelligence](/content/explore?query=Nature%20Machine%20Intelligence&refine=journal/index.html)

[Biology\\
\\
28 \| Jan \| 2026\\
\\
Deconstruction of a memory engram reveals distinct ensembles recruited at learning\\
\\
This code capsule contains all the necessary codes to obtain the correlation results of the miniscope part of the article (See figure4J). Cumulative distribution figures are saved under /results/figures/XX\_cumsum.pdf; while a visual representation of the associated one-way ANOVA is shown in XX\_stats.pdf. the p value in the title refers to the p value of the zoomed-in bar graphs of figure 4J. Complete statistics of ANOVAs are stored under /results/anova\_stats.mat. All the middle steps are stored in their associated /results/ folders for each animal and experiment, e.g. index of shock cells, freezing cells, etc.\\
\\
Clément Pouget](/content/explore/ea8e6639-f0b1-4bd8-9ff7-04f3d06bcb62?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4121623/tree/v1/index.html)

published in [Nature Neuroscience](/content/explore?query=Nature%20Neuroscience&refine=journal/index.html)

[Biology\\
\\
28 \| Jan \| 2026\\
\\
eQTL in diseased colon tissue identifies novel potential target genes associated with IBD\\
\\
IBD colon eQTL\\
\\
Nina Nishiyama](/content/explore/3d8a6465-6614-4043-887f-b4788af18316?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1077277/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Physics\\
\\
28 \| Jan \| 2026\\
\\
code for "Emergent giant topological Hall effect in twisted Fe3GeTe2 metallic system"\\
\\
Code for "Emergent giant topological Hall effect in twisted Fe3GeTe2 metallic system"\\
\\
Yuhang Li](/content/explore/04a765dd-c0ff-4e52-9a25-a010f95df4bd?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2336028/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
27 \| Jan \| 2026\\
\\
Algorithmic Design of Realizable Low-Order IIR Filters for Phase Matching\\
\\
Based on a manuscript being submitted for publication. Abstract is as follows:\\
Designing digital filters with desired phase responses\\
at selected frequencies is essential in several communication\\
and control applications. While phase compensation is generally\\
achieved using all-pass filters, the all-pass constraint limits the\\
range of realizable stable, causal filters that match the desired\\
phase characteristic. In this paper, we present a method for\\
designing real, stable discrete-time filters that match phase\\
constraints at selected frequencies. Our approach enables control\\
over the phase response without affecting stability and with\\
low complexity. Specifically, we construct an nth-order filter\\
to match phase constraints at n distinct frequencies (for even\\
n), achieving exact phase interpolation with a low filter order.\\
The derived methodology cascades suitably designed second-\\
order sections, referred to as filter blocks, and uses a fixed-point\\
iteration scheme to meet the phase specifications, with tunable\\
design parameters that enable explicit control over the margin\\
of stability. Simulations confirm the utility of our method for\\
obtaining stable filters that satisfy phase specifications even where\\
comparable methods, such as all-pass designs, fail.\\
\\
Rishabh Shetty, Kumar Appaiah & Vivek Natarajan](/content/explore/1938a270-e0b2-4932-80d4-8de59bf006c3?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6735133/tree/v1/index.html)

[Mathematics\\
\\
27 \| Jan \| 2026\\
\\
Almost Exact Graph Matchings\\
\\
Code for the paper Exact recovery in almost fully seeded graph matching submitted to IEEE Pattern Analysis.\\
\\
Michael Nisenzon & Nicolas Fraiman](/content/explore/489abc8e-112f-42a2-8707-a805e1dc9abd?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1032313/tree/v1/index.html)

[Computer Science\\
\\
10 \| Feb \| 2026\\
\\
Physics-Informed Priors for Bayesian Neural Network with Uncertainty Quantification in Battery Pack Health Assessment\\
\\
Lithium-ion battery health-state estimation is a key component of battery management systems, with current methods primarily relying on model-based or data-driven approaches. This article proposes a physics-informed prior for Bayesian neural networks framework for early prediction of battery pack capacity degradation with uncertainty quantification. To address data scarcity and complex degradation dynamics at the pack level, domain knowledge is incorporated directly in function space by constructing physics-informed priors using Gaussian process regression. These priors encode electrochemical and thermal degradation characteristics through kernel designs and are transferred to a Bayesian convolutional neural network via function-space variational inference. Unlike conventional parameter-space priors, the proposed formulation enables physically meaningful regularization of the learned degradation functions. The Bayesian neural network further provides uncertainty quantification by disentangling predictive uncertainty into epistemic and aleatoric using the law of total variance, yielding calibrated confidence estimates for mid- and long-term predictions from early-cycle data. The proposed approach is evaluated on NASA lithium-ion battery pack degradation datasets collected under diverse operating conditions and compared against representative physics-aware Bayesian baselines. Experimental results demonstrate improvements in predictive accuracy and uncertainty calibration while maintaining physically consistent degradation trends.\\
\\
Jokin](/content/explore/cd7c9d90-69a6-4bcf-85b5-75dec3d3135e?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6780240/tree/v1/index.html)

[Bioinformatics\\
\\
26 \| Jan \| 2026\\
\\
Use of ctDNA in Older Women with ER+ Breast Cancer to Facilitate Surgical De-Escalation: A Prospective, Hybrid-Decentralized Trial with Correlative Studies\\
\\
Code capsule for "Use of ctDNA in Older Women with ER+ Breast Cancer to Facilitate Surgical De-Escalation: A Prospective, Hybrid-Decentralized Trial with Correlative Studies" by Carleton et.al, 2026.\\
\\
Neil Carleton et al.](/content/explore/6055a39b-77bc-45e4-95b5-1a6d1f09959a?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5420235/tree/v1/index.html)

published in [Clinical Cancer Research](/content/explore?query=Clinical%20Cancer%20Research&refine=journal/index.html)

[Physics\\
\\
26 \| Jan \| 2026\\
\\
No anti-solar: Solar-type stars always have solar-like differential rotation\\
\\
For more than 45 years, scientists have believed in the existence of two classes of differential rotation (DR), solar-like and anti-solar.\\
The Sun rotates differentially with a fast equator and slow poles, called solar-like DR. Theoretical studies suggest that this DR becomes anti-solar in slowly rotating stars, where the poles rotate faster. The DR topology significantly affects the stellar magnetic activity and long-time stellar evolution. According to recently reported observations, however, there are inconsistencies with this solar-like and anti-solar idea in several aspects.\\
We carry out unprecedentedly high-resolution magnetohydrodynamic simulations for the slowly rotating solar-type stars. The simulated solar-type stars always show solar-like DR even with significantly slow rotation rates in our simulations. Here we show that no anti-solar DR is achieved. The strong magnetic field maintains the solar-like DR in all cases. The rotation has a less significant influence on the magnetic field strength than that on the turbulence anisotropy.\\
Our results show that the magnetic field monotonically decreases over the stellar lifetime, indicating a weakening of the magnetic braking. This trend is also consistent with recent observational results of the stellar rotation evolution.\\
\\
Hideyuki Hotta & Yoshiki Hatta](/content/explore/992fe518-b66e-4f36-9dd5-515eaf4b91b5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2529293/tree/v1/index.html)

published in [Nature Astronomy](/content/explore?query=Nature%20Astronomy&refine=journal/index.html)

[Bioinformatics\\
\\
26 \| Jan \| 2026\\
\\
stVCR: Spatiotemporal dynamics of single cells\\
\\
This is the code implementation of the paper "stVCR: Spatiotemporal dynamics of single cells".\\
\\
Qiangwei Peng](/content/explore/8c082f44-190b-4f9b-96b1-f6f91d5491ed?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7965529/tree/v1/index.html)

published in [Nature Methods](/content/explore?query=Nature%20Methods&refine=journal/index.html)

[Engineering\\
\\
1 \| Mar \| 2026\\
\\
A Discrete Data-Driven Transfer Function Identification Vector Fitting Technique with FIR Correction for DC-DC Converters\\
\\
This project provides an identification algorithm for transfer functions (TF). FIR-ZD-VF is an FIR-corrected Z-domain data-driven TF vector fitting technique. The algorithm is intended for embedded adaptive control systems, in particular DC/DC converters.\\
\\
- A Discrete Data-Driven Transfer Function Identification Vector Fitting Technique with FIR Correction for DC-DC Converters\\
- Journal IEEE Transactions on Power Delivery \\
- Submitted On 27 February 2026\\
- Regular Paper\\
\\
Aleksei Chernyshov, Mikhail Pugach, Nikolay Kalugin & Federico Martin Ibanez](/content/explore/c2731674-86d9-4405-9488-fa0475c3a901?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5450683/tree/v2/index.html)

[Computer Science\\
\\
26 \| Jan \| 2026\\
\\
Policy-Driven DSCP Edge Enforcement for DOCSIS Upstream: CSIT Reproducibility Artifact\\
\\
This capsule is the reproducibility artifact for the manuscript **“Policy-Driven Edge Enforcement for DOCSIS Upstream in HFC Networks”** (IEEE Communications Magazine, under review). It enables practitioners to reproduce the experimental workflow used to validate **DSCP-based, policy-driven per-class enforcement** (e.g., shaping/policing/controlled degradation) at the access edge using **VPP** and the **FD.io CSIT** methodology.\\
\\
**What runs in Code Ocean**\\
\\
- Generate a CSIT-compatible topology file: `generated/topology.yaml` from environment variables.\\
- Clone baseline repositories:\\
  - CSIT: `https://github.com/FDio/csit` (ref: `master`)\\
  - VPP: `https://github.com/FDio/vpp` (ref: `master`)\\
- Apply `code/patches/csit.patch` to CSIT (optional `code/patches/vpp.patch`).\\
- Run CSIT autogeneration: `tox -e autogen`.\\
- Package a lab-ready bundle:\\
  - `results/csit_bundle.tar.gz`\\
  - `results/manifest.json`\\
  - `results/RUN_ON_LAB.txt`\\
\\
**What runs on the user’s lab orchestrator (hardware-in-the-loop)** \\
\\
The actual CSIT execution requires an external orchestrator with SSH access to **two physical DUTs** interconnected with supported NICs and configured for DPDK.\\
The canonical command template is:\\
`robot -v TOPOLOGY_PATH:<path> -L TRACE -s tests.vpp.dscp -i <tag>`\\
\\
**Inputs**\\
\\
- Lab-specific environment variables (DUT hostnames, NIC model/driver, PCI BDF, IP addressing, SSH credentials) are documented in `.env.example`.\\
\\
**Outputs**\\
\\
- `generated/topology.yaml`\\
- `results/csit_bundle.tar.gz` (one-command lab execution bundle)\\
- `results/manifest.json`\\
- `results/RUN_ON_LAB.txt`\\
\\
See `code/README` for full testbed requirements and step-by-step usage.\\
\\
Ivan Ivanets](/content/explore/35807129-f470-42da-86b4-bb5e0e3b6452?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0826802/tree/v1/index.html)

[Medical Sciences\\
\\
24 \| Jan \| 2026\\
\\
Early Diagnosis of Pancreatic Ductal Adenocarcinoma by Signal-Enhanced Lateral Flow Immunoassay: SELFI\\
\\
Pancreatic ductal adenocarcinoma (PDAC) is linked to high incidence and mortality rates because it is often detected in later stages, when the prognosis is poor. However, the current state-of-the-art methods for diagnosing early PDAC tend to be invasive, time-consuming, and unreliable, primarily due to the difficulties associated with the early detection of pancreatic cancers. Here, we report a quick and sensitive method for the early diagnosis of PDAC using a signal-enhanced lateral flow immunoassay called SELFI. We developed SELFI, which can generate a strong colorimetric signal through multiple hotspots formed by plasmonic gold nanoparticles (AuNPs) assembled on a silica bead. Our SELFI assay achieved a 28-fold increase in the limit of detection compared to conventional lateral flow immunoassays using 20 nm AuNPs, providing results within 15 min. We demonstrated that SELFI can be utilized for the early diagnosis of PDAC, as indicated by a receiver operating characteristic curve and a larger area under the curve compared to the enzyme-linked immunosorbent assay (\*\*\*\*P < 0.0001). SELFI's effective diagnostic features could enhance the timely identification of PDAC and may also serve in the early diagnosis of a range of other diseases.\\
\\
Sohyeon Jang et al.](/content/explore/90b34f8d-18fa-4897-8b02-23aa9be23c04?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7919487/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Physics\\
\\
23 \| Jan \| 2026\\
\\
Continuously trapped matter-wave interferometry in magic Floquet-Bloch band structures](/content/explore/1713a82a-f0aa-4dda-b94a-a58936debc26?page=2&filter=all/index.html)

[This capsule provides the calculations of magic Floquet-Bloch band structures. It is based on a package we developed named](/content/explore/1713a82a-f0aa-4dda-b94a-a58936debc26?page=2&filter=all/index.html) [MuscleMuseum](https://github.com/XiaoCasd/MuscleMuseum) which computes the lattice properties. Simply run `main.m` to test the functionality.

The `plotFloquetBand` script computes and plots the Floquet-Bloch P-D hybridized bands, corresponding to Figure 1 of the main text. The produced figure is saved in `/results`

The `findMagicDepthStatic` and `findMagicDepthDressed` scripts compute magic band structures assuming static and dressed bands, respectively. They produce `.mat` files saved in `/results`. Each `.mat` contains variables including:

- loopSizeTheory: loop size in 2ℏkL2\\hbar k\_L2ℏkL​
- magicDepthTheory: magic depth in ERE\_RER​
- modDepth: relative modulation depth
- Frequency: modulation frequency in Hz
- fringeFrequency: fringe frequency in Hz

`findMagicDepthDressed` computes two extra quantities:

- toleranceTheory: tolerance against depth fluctuations, in ERE\_RER​
- toleranceTheoryRelative : relative tolerance again depth fluctuations

Note: in `findMagicDepthStatic` and `findMagicDepthDressed`, the variable `nl` can be modified to set how fine we scan across the loop size range. The `divisor` parameter can be changed to modify the applied force, given by F=Mg/divisorF=Mg/divisorF=Mg/divisor.

Xiao Chai et al.

[Open Capsule](/content/capsule/8593166/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Computer Science\\
\\
22 \| Jan \| 2026\\
\\
SB-Net: Software for Classifying Botnet Attacks to Prevent Comprehensive Infection\\
\\
Botnet detection\\
\\
Reynandriel Pramas Thandya, Algof Kristian Zega, Irsyad Fikriansyah Ramadhan & Tohari Ahmad](/content/explore/9190d97c-62df-43fb-bcd9-414f6202fdf2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6872668/tree/v1/index.html)

[Computer Science\\
\\
22 \| Jan \| 2026\\
\\
Optimizing Sparse Iterative Solvers with Stream-Based GPU Parallelism: A Case Study on BiCGStab\\
\\
The zip file contains the code files and the benchmarking scripts.\\
\\
Ayaz ul Hassan Khan](/content/explore/c94b58be-e75c-4702-b30d-716b6831be97?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3337730/tree/v1/index.html)

[Chemistry\\
\\
21 \| Jan \| 2026\\
\\
Realistic atomic model for amorphous nanoporous carbons\\
\\
Amorphous porous carbons have been widely used as electrodes for energy storage. However, due to their structural heterogeneity and complex pore topology, the absence of reliable atomic carbon models hinders understanding energy storage mechanisms through molecular simulations. Here, we developed a modelling approach capturing the rich experimental information of small angle X-ray scattering, gas adsorption for three-dimensional morphology (mGRF) construction\\
\\
Guang Feng & Jiaxing Peng](/content/explore/937a22fe-dcf2-40e0-9c24-a474c46e8d43?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5025667/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
20 \| Jan \| 2026\\
\\
The use of multi-sensor drone data for the development and validation of methods to track and characterize marine animals\\
\\
Low cost, unmodified, commercially available drones can provide an effective platform for the study and characterization of marine megafauna. We present methods which utilize video and flight data to allow for both the continuous tracking of animals and the determination of animal lengths across a range of flight parameters. We also provide a thorough estimation of error in animal position and length measurements while at the same time introducing methods to correct for errors in reported aircraft altitude and heading. Methods are validated using both ground-based markers and tracking data from free swimming white sharks which includes the simultaneous tracking of individual sharks by two drones as the aircraft undergo changes in altitude, gimbal angle, heading and position. The resultant tracks are seen to be highly congruent (mean distance between measured positions - 4.3 m (95% CI 0 to 10)) and length measurements demonstrate a high level of precision (95% CI −8 to 8%) with accuracy confirmed using ground-based markers (mean error - 0.3% (95% CI −4.8 to 4.8%)). Results demonstrate the effectiveness of these methods across a range of flight conditions encountered in the field. The methods introduced allow flexibility in data capture while still providing accurate information, with the potential to both expand the use of and enhance the value of drone-based data for the quantification of animal behaviors and characteristics.'\\
\\
Kristian J Sexton et al.](/content/explore/7161063e-4f55-407a-8846-4c77f574a3ba?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0294361/tree/v1/index.html)

[Associated article](https://doi.org/10.1038/s41598-025-31975-2) published in [Scientific Reports](/content/explore?query=Scientific%20Reports&refine=journal/index.html)

[Computer Science\\
\\
20 \| Jan \| 2026\\
\\
EASELAN: An Open-Source Framework for Multimodal Biosignal Annotation and Data Management](/content/explore/855ca74f-5971-43a2-b61a-84cb130f4317?page=2&filter=all/index.html)

[Recent advancements in machine learning and adaptive cognitive systems frequently rely on large and richly annotated multimodal data like fusion models that incorporate multiple biosignals as well as traditional audiovisual channels. This paper introduces the EASELAN annotation framework to improve annotation workflows that address the complexity of multimodal and biosignal datasets. EASELAN builds on ELAN by adding new components to support all stages of the annotation pipeline, including preparation of annotation files, setting up additional channels, version control with GitHub, and simplified post-processing. The EASELAN workflow integrates biosignals and facilitates rich annotations to be exported for further analyses and machine learning-supported model training. We successfully applied EASELAN to a high-dimensional biosignals database on human everyday table setting for cognitive robotics, and discuss the opportunities, limitations, and lessons learned. The code of EASELAN (](/content/explore/855ca74f-5971-43a2-b61a-84cb130f4317?page=2&filter=all/index.html) [https://github.com/cognitive-systems-lab/easelan](https://github.com/cognitive-systems-lab/easelan)) and the EASELAN-supported fully annotated Table Setting Database are both publicly available.

Rathi Adarshi Rammohan, Moritz Meier, Dennis Küster & Tanja Schultz

[Open Capsule](/content/capsule/8137497/tree/v1/index.html)

[Earth Sciences\\
\\
20 \| Jan \| 2026\\
\\
Collection: Datasets From Real-Time In-Situ Soil Monitoring for Agriculture 2025](/content/explore/a6f61ccb-d208-44ad-97e6-c91dcb91958a?page=2&filter=all/index.html)

[This capsule accompanies the IEEE article Collection. Datasets From Real-Time In-Situ Soil Monitoring for Agriculture 2025 by Kayla R. Moore, Taras Lychuk, Heather Mcnairn, Xiaoyuan Geng, Aston Chipanshi, E. Rotimi Ojo and 15 additional co-authors at](/content/explore/a6f61ccb-d208-44ad-97e6-c91dcb91958a?page=2&filter=all/index.html) [https://ieeexplore.ieee.org/document/11177568](https://ieeexplore.ieee.org/document/11177568). \\n\\nIt provides a minimal but complete example of the RISMA Calibration and Quality Control Process workflow. The full RISMA system processes soil moisture data from **36 monitoring stations** across three Canadian agricultural regions **24 in Manitoba** (plus two meteorological stations), **8 in Ontario**, and **4 in Saskatchewan**. The code complies data from soil moisture and weather sensors in the field. Site-specific calibration equations are applied for soil moisture sensors. It then quality controls the data, applying flags for data that are suspected to be erroneous. Full details on the calibration process and flags are provided in the associated publication.\\n\\nAlthough the full system includes 36 stations, SK3 is provided as a representative dataset to demonstrate the data processing pipeline, metadata usage, and output structure."

Belinda Bence et al.

[Open Capsule](/content/capsule/0261547/tree/v1/index.html)

[Associated article](https://doi.org/10.1109/ieeedata.2025.3612373) published in [IEEE Data Descriptions](/content/explore?query=IEEE%20Data%20Descriptions&refine=journal/index.html)

[Biology\\
\\
20 \| Jan \| 2026\\
\\
SpatiotemporalMotifs.jl: Analyzing nested spatiotemporal waves in mouse Neuropixels data\\
\\
Julia code for reproducing the results of " _Nested spatiotemporal theta-gamma waves organize hierarchical visual processing_".\\
In this work, we analyzed mouse Neuropixels data to reveal a cross-scale principle of spatiotemporal dynamics involving low-frequency θ, high-frequency γ, and neuronal spiking that coordinate translaminar and hierarchical visual processing.\\
Detailed instructions for reproducing calculations and figures can be found in `code/SpatiotemporalMotifs.jl/README.md`.\\
\\
Brendan Harris & Pulin Gong](/content/explore/91bf4eb6-2352-4bcb-931f-e801c05176dd?page=2&filter=all/index.html) [Open Capsule](/content/capsule/4585157/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Chemistry\\
\\
20 \| Jan \| 2026\\
\\
Catalytic Proximal Protein Oligomerization (CaPPO) as a Novel Anti-Tumor Strategy Targeting WDR5\\
\\
Code for extracting Peak Current and Dwell Time from every events of nanopore sensing in the article of "Catalytic Proximal Protein Oligomerization (CaPPO) as a Novel Anti-Tumor Strategy Targeting WDR5".\\
\\
Yizheng Fang et al.](/content/explore/59a8cfcc-fa95-4b9e-a6e5-6e9e1d92f4d5?page=2&filter=all/index.html) [Open Capsule](/content/capsule/3397769/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Chemistry\\
\\
19 \| Mar \| 2026\\
\\
Gallium in Liquid State Shows Nuclease-Mimicking Activity\\
\\
This repository contains all code and data necessary to reproduce the ONT DNA sequencing results presented in the paper titled "Gallium in Liquid State Shows Nuclease-Mimicking Activity".\\
\\
Li Liu](/content/explore/4a7421e5-5288-4550-bb35-992d43c44caa?page=2&filter=all/index.html) [Open Capsule](/content/capsule/0837350/tree/v3/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Engineering\\
\\
26 \| Jan \| 2026\\
\\
IMPACT: A Toolchain to Simulate and Deploy Nonlinear Model Predictive Control. Documentation example\\
\\
This code executes the example provided in the Impact toolchain documentation, implemented in Python but also available in Matlab. It first defines the system model, followed by the formulation of the optimal control problem OCP. A transcription method and a solver are then selected, in this case, the Fatrop solver. The problem is solved, and the results are plotted. Next, the same process is executed in a loop to emulate Model Predictive Control MPC, and the corresponding results are also visualized. Finally, the problem is exported and the associated artifacts are generated, including C functions, a Simulink block, and a ROS 2 node.\\
\\
Alvaro Florez et al.](/content/explore/ea91aac5-30a1-418b-b826-fae621e706fb?page=2&filter=all/index.html) [Open Capsule](/content/capsule/9887084/tree/v2/index.html)

[Engineering\\
\\
19 \| Jan \| 2026\\
\\
Bioinspired spiking architecture enables energy constrained touch encoding\\
\\
This capsule enables the reproduction of the main results presented in the paper: "Bioinspired Spiking Architecture Enables Energy-Constrained Touch Encoding".\\
\\
Andrea Ortone et al.](/content/explore/88147aef-99a1-438f-b659-e093da9b9277?page=2&filter=all/index.html) [Open Capsule](/content/capsule/1018585/tree/v1/index.html)

published in [Nature Communications](/content/explore?query=Nature%20Communications&refine=journal/index.html)

[Social Sciences\\
\\
29 \| Jan \| 2026\\
\\
A Socio-Ecological Analysis of Factors Influencing Perceived Value of Multi-Cancer Early Detection Testing in the United States\\
\\
Using a socio-ecological framework, we analyzed nationally representative data from the 2024 Health Information National Trends Survey to examine predictors of perceived MCED value across proximal, intermediate and distal levels. Survey-weighted logistic regressions estimated adjusted odds ratios (aOR) with 95% confidence intervals.\\
\\
Chuqiao Wang](/content/explore/d331ef2b-6953-4b9c-98c6-f680fc91e19c?page=2&filter=all/index.html) [Open Capsule](/content/capsule/5312089/tree/v2/index.html)

[Engineering\\
\\
15 \| Jan \| 2026\\
\\
Adaptive Neuron Selective Aggregator (ANSA): An Experimental Comparison of Federated Learning Robustness Under Data Poisoning\\
\\
This code implements a unified experimental framework to compare Centralized Machine Learning, Standard Federated Learning (FedAvg), LFighter, and the proposed ANSA defense under controlled label-flipping attack scenarios.\\
\\
The framework is designed to ensure fairness, reproducibility, and consistency across all methods by enforcing identical data partitions, malicious client selections, and attack configurations.\\
\\
The MNIST dataset is used as the benchmark. From the original training set, 50,000 samples are selected and partitioned equally among N clients, where the number of clients is specified by the user. Each client holds a disjoint local dataset, reflecting a standard federated learning setting. The MNIST test set is split into per-client validation sets, while the full test set is used to evaluate the global model. All data splits are deterministic under a fixed random seed.\\
\\
Seven attack scenarios are evaluated. Scenario S1 contains no attack. Scenarios S2–S4 involve fewer than 50% malicious clients, while Scenarios S5–S7 involve more than 50% malicious clients. For each scenario, malicious clients are selected randomly but deterministically, and the same malicious client set is reused across all methods to ensure a fair comparison.\\
\\
The attack model is label-flipping poisoning, where specific source classes are flipped to a target class (class 2). Attack strength increases across scenarios by flipping additional classes: scenarios S2 and S5 flip class 7; S3 and S6 flip classes 7 and 3; and S4 and S7 flip classes 7, 3, and 9. These operations are applied only to the malicious clients’ local training data.\\
\\
Four learning paradigms are compared. In Centralized Learning, a single model is trained on the union of all client data, including poisoned samples. FedAvg performs standard federated aggregation without defenses. LFighter detects and excludes suspicious client updates based on clustering of model updates. ANSA selectively filters suspicious neurons in client updates, mitigating poisoning while preserving benign information.\\
\\
All federated methods operate in communication rounds with multiple local epochs per round. Centralized learning is trained for an equivalent total number of epochs. For each scenario and method, the framework records per-client validation accuracy, global accuracy per round, and final global test accuracy.\\
\\
For each scenario, a comparison plot is generated showing global model accuracy per round for all methods, with centralized results aligned to the federated rounds. All global accuracy curves are plotted in black for visual consistency. The framework is fully reproducible and controlled through the number of clients and a single random seed.\\
\\
Rishit Prajapati & Hong Liu](/content/explore/18b035ef-6af4-4010-bc0b-48c99946b7c2?page=2&filter=all/index.html) [Open Capsule](/content/capsule/8955838/tree/v1/index.html)

[Engineering\\
\\
15 \| Jan \| 2026\\
\\
Transcoding-Free Voice Services over GEO Satellites: A Comparative Guide to Heuristic, Learning, and Optimal Approaches\\
\\
This repository reproduces the numerical evaluation of transcoding-free GEO voice call setup optimization. It models stochastic GEO impairments (delay, jitter, packet loss) and varying GEO satellite transmission rates (0.8-4.0 Kbps). via Monte Carlo sampling and compares Enhanced ROHC, Centralized Linear Programming, and Reinforcement Learning approaches for header and IMS signaling efficiency. The scripts output the key plots used in the manuscript, including reduction ratios, end-to-end signaling size, and runtime trends\\
\\
Ahmed Mohammed MIkaeil](/content/explore/cb8fefc3-2ecf-45ca-8231-b006ebc29fc0?page=2&filter=all/index.html) [Open Capsule](/content/capsule/6093406/tree/v1/index.html)

[Computer Science\\
\\
14 \| Jan \| 2026\\
\\
Reprogramming LLM Semantics: a Symbolic Attack using Words Reframing\\
\\
Disclaimer: The data contains potentially offensive and harmful text.\\
\\
This capsule contains files for the "Reprogramming LLM Semantics: a Symbolic Attack using Words Reframing" paper:\\
\\
- llamator-EG-attack.ipynb - LLM Security Pipeline with custom attack running\\
- asr.ipynb - Notebook for ASR counting\\
\\
Large language models (LLMs) are widely used because they are simple and powerful, yet safe deployment remains difficult. Though post-generation moderation and alignment have improved, adversarial prompting is still a major weakness: many defenses detect only surface patterns and ignore deeper semantic structure. We present a black-box jailbreak that exploits semantic reframing. Without accessing weights, logits, or internal prompts, we add a symbolic abstraction layer that keeps unsafe words’ surface form but redefines their meaning as abstract entities within the dialogue. This subtle reinterpretation lets LLMs process and generate content consistent with the new semantics, bypassing refusals. We evaluated the method on several leading models from different vendors using automated red-team tools on a subset of HarmBench across multiple harm categories. The attack achieves over 90% refusal-attack success and up to 78% ASR under the benchmark protocol, exposing a previously overlooked vulnerability tied to semantic reinterpretation. Our analysis also highlights factors that affect model resistance, offering insights for strengthening alignment against misuse such as disinformation or inappropriate content.\\
\\
RUDOLF S. STASILOVICH, TIMUR D. NIZAMOV, EVGENIY S. KOKUYKIN & DMITRY S. BOTOV](/content/explore/98ca9ac6-9480-42bd-8f37-88fc6db21846?page=2&filter=all/index.html) [Open Capsule](/content/capsule/7845102/tree/v1/index.html)

[Social Sciences\\
\\
15 \| Jan \| 2026\\
\\
From Waste to Fuel: Uncovering the Drivers of Garbage and Plastic Burning in Northwest Nigeria\\
\\
Anecdotal evidence indicates that energy-poor households in regions lacking formal waste management systems often resort to burning garbage, including plastic, to fulfill their energy requirements and manage waste. This study utilizes the Multi-Tier Framework (MTF) dataset from Northwest Nigeria to examine the prevalence of garbage and plastic burning, along with the underlying reasons and methods households employ for this practice. Analyzing data from 3,669 households, findings show that up to 58% of households in certain local government areas (LGAs) burn garbage including plastic as fuel. Factors driving this behavior extend beyond economic constraints, notably increased plastic waste, inadequate waste management infrastructure, and declining availability of traditional fuels. In some instances, external influences significantly motivate the use of polluting fuels. This localized investigation underscores the emerging challenge of plastic burning in impoverished communities in the Global South, where rising plastic pollution, diminishing biomass resources, and the absence of formal waste services converge, exacerbating health and environmental risks.\\
\\
Bishal Bharadwaj, Tara Gates, Ian Gates & Srijani Deb](/content/explore/dce6c479-fdd7-45b0-8aa5-72f81148771d?page=2&filter=all/index.html) [Open Capsule](/content/capsule/2413685/tree/v2/index.html)

[Biology\\
\\
12 \| Jan \| 2026\\
\\
Benchmarking PRS methods for risk prediction of 36 complex traits in UK Biobank by establishment of a large-scale PRS computation platform PRS-hub](/content/explore/cac63269-145b-4733-b1ff-5a82dfd52987?page=2&filter=all/index.html)

[**PRS-hub\_offline** \\
PRS-hub is a platform for batch offline calculation of various Polygenic Risk Score (PRS) methods, built on WDL, allowing for local or server-based execution. Below are detailed instructions for installation, dependencies, and usage.\\
**Installation**\\
\\
1. Clone this repository:\\
\\
\\
\\
bash\\
\\
Copy\\
\\
```bash\\
git clone https://github.com/yourusername/PRS-hub.git\\
```\\
\\
2. Set the `PRSHUB_PATH` environment variable in your `.bashrc` file to quickly call scripts from the installation path:\\
\\
\\
\\
bash\\
\\
Copy\\
\\
```bash\\
export PRSHUB_PATH=`pwd`/PRS-hub_offline\\
```\\
\\
3. Save and source the `.bashrc` file to apply changes:\\
\\
\\
\\
bash\\
\\
Copy\\
\\
```bash\\
source ~/.bashrc\\
```\\
\\
\\
If the installation path changes, simply update the `PRSHUB_PATH` value accordingly.\\
**Dependencies** \\
To ensure the platform runs correctly, please install the following software and dependencies:](/content/explore/cac63269-145b-4733-b1ff-5a82dfd52987?page=2&filter=all/index.html)

2. [**Cromwell**: This platform is based on WDL (Workflow Description Language), so Cromwell is required to execute WDL scripts. You can download Cromwell and refer to the installation instructions on the](/content/explore/cac63269-145b-4733-b1ff-5a82dfd52987?page=2&filter=all/index.html) [Cromwell GitHub page](https://github.com/broadinstitute/cromwell).

[**R and Python**:\\
\\
   - **R** and **Python** need to be included in the `PATH` to allow direct usage of the `Rscript` and `python` commands. If not installed, they can be added via Conda, and the Conda environment path can be appended to `.bashrc`.\\
     \\
     For example, install R and Python using Conda:\\
     \\
     \\
     \\
     bash\\
     \\
     Copy\\
     \\
     ```bash\\
     conda install -c conda-forge r-base python\\
     ```\\
**R Packages**: Install the following R packages to ensure compatibility with the environment.\\
\\
   - Use the following command to install the required R packages:\\
     \\
     \\
     \\
     R\\
     \\
     Copy\\
     \\
     ```r\\
     install.packages(c("pacman", "dplyr", "docopt", "rio", "data.table", "magrittr", "bigsnpr"))\\
     ```](/content/explore/cac63269-145b-4733-b1ff-5a82dfd52987?page=2&filter=all/index.html) 4. [**Python Packages**:](/content/explore/cac63269-145b-4733-b1ff-5a82dfd52987?page=2&filter=all/index.html)

- [**scipy**: Required for scientific computing. For details, refer to the](/content/explore/cac63269-145b-4733-b1ff-5a82dfd52987?page=2&filter=all/index.html) [Scipy official website](https://www.scipy.org/).
   - **h5py**: Needed for handling HDF5 file formats. For details, refer to the [H5py official website](https://www.h5py.org/).

Install the Python packages with:

bash

Copy

```bash
pip install scipy h5py

Usage

1. To execute a specific algorithm WDL file using Cromwell, use the following command, replacing <algorithm_name> with the name of the desired algorithm:

bash

Copy

java -jar /path/to/cromwell.jar run $PRSHUB_PATH/wdl/<algorithmname>.wdl --inputs /path/to/config_file

• /path/to/cromwell.jar is the path to the Cromwell JAR file.
  • $PRSHUB_PATH/wdl/<algorithm_name>.wdl is the path to the WDL file for the chosen algorithm.
  • --inputs /path/to/config_file specifies the path to the configuration file, which contains the parameter settings for the algorithm.

2. Each algorithm's supported configuration file parameters can be found in the config_example folder. Adjust the parameters in the configuration file according to the requirements of each algorithm.

Thank you for using PRS-hub!

Xingyu Chen & Fei Wang

Open Capsule

published in Nature Communications

Social Sciences\ \ 2 | Feb | 2026\ \ Adaptive trophic dismantling exposes multi-scale structural cascades in directed networks\ \ Adaptive trophic dismantling reveals scale-free structural avalanches in directed networks\ \ Jiawen Hu & Xueming Liu Open Capsule

Engineering\ \ 9 | Jan | 2026\ \ China’s offshore wind power potential revisited: a realistic appraisal sug-gests substantially lower estimates than precedent studies\ \ This repository contains code for assessing offshore wind energy potential at the farm level scale for China. The workflow requires running the scripts in the following order: AEP-Wake loss.py, Main-calculate.py, and Farm-LCOE.py. See the readme.md file for a detailed description.\ \ Shiwei Xu Open Capsule

published in Nature Communications

Social Sciences\ \ 8 | Jan | 2026\ \ Personal appeals to male investors can backfire\ \ Women’s reproductive-health ventures face persistent funding gaps, partly because most investors are men who favor ventures aligned with their interests. Research suggests that engaging broader social roles and identities can motivate support for initiatives that do not directly benefit the decision-maker. We therefore test whether relational appeals (asking potential backers to consider the women in their lives) increase male interest. We report on two preregistered field experiments embedded in real fundraising campaigns. One venture, related to a period-care product (N=34,943), solicited funding via emails advertising an equity crowdfunding campaign. The second, related to a contraceptive pill (N=1,805,472), solicited funding via Facebook advertisements. Investor interest was measured based on email or ad click-through, respectively. Across both experiments, relational appeals reduced men’s interest relative to financially focused market-attractiveness appeals (e.g., highlighting the market opportunity). In contrast, though women's interest was systematically higher, it did not differ across appeal types.\ \ Sofia Bapna & Gordon Burtch Open Capsule

Medical Sciences\ \ 8 | Jan | 2026\ \ Targeted innate immune inhibition therapy for treatment of recurrent acute cystitis\ \ This capsule contains the data from a randomized phase 2 trial comparing the IL-1 receptor antagonist anakinra with nitrofurantoin in women with recurrent cystitis. It includes Acute Cystitis Symptom Scores, quality-of-life measures, white blood cell counts, and differential counts. The provided R script describes the analysis code for ordinal logistic mixed-effects and ART-ANOVA models and post-hoc comparisons. Associated clinical trial identifiers: DRKS00025964 and EudraCT 2019-004209-28.\ \ Ines Ambite Open Capsule

published in Nature Microbiology

Bioinformatics\ \ 8 | Jan | 2026\ \ Deep learning-based semantic matching of cis-regulatory DNA sequences facilitates the prediction of gene function\ \ PhytoBabel \ \ The rich information encoded in cis-regulatory DNA sequences has not been fully exploited for gene function prediction in reverse genetics. Here we show that orthologous cis-regulatory sequences that diverged approximately 160 million years ago share little sequence similarity, yet remarkably retain semantic similarity that can be effectively captured by a deep learning model, PhytoBabel. Although trained solely on orthologous cis-regulatory sequence pairs from 15 angiosperms, PhytoBabel implicitly learned spatio-temporal gene expression patterns, conserved non-coding sequences, semantically similar fragments, and phylogenetic relationships among species. Furthermore, PhytoBabel enables the discovery of evolutionarily unrelated but semantically similar cis-regulatory sequences, facilitating the identification of novel genes with functions of interest. As a proof-of-concept, we identified in maize new somatic embryogenesis-related morphogenic regulators exhibiting semantic similarity to known Arabidopsis morphogenic regulators. By bridging the gap in the cis-regulatory sequence → semantics → gene function information chain, PhytoBabel provides a valuable tool for gene function prediction in reverse genetics.\ \ Semantic similarity of cis-regulatory DNA sequences prediction \ \ python model_predict.py -r ath_ref_gene.csv -q zma_query_gene.csv -m PhytoBabel_model -s pred_out.csv\ \ Parameters\ \ -r : Cis-regulatory sequences of reference gene file\ \ -q : Cis-regulatory sequences of query gene file\ \ -m : The directory containing all models for prediction\ \ -s : Semantic similarity prediction results file path\ \ -g : Specify the gpu usage\ \ The scripts used for article analysis can all be found in ‘code used in the manuscript’\ \ tianyi Li Open Capsule

published in Nature Plants

Biology\ \ 7 | Jan | 2026\ \ Polyauxic Modeling Platform (v1.0.0)\ \ Polyauxic Modeling Plataform is an open-source Python package designed for the robust kinetic modeling of mono- and polyauxic microbial growth. Addressing the limitations of standard single-phase analysis, this software implements semi-mechanistic reformulations of Boltzmann and Gompertz equations to extract biologically interpretable parameters from complex datasets. The computational workflow integrates a hybrid optimization pipeline, combining Differential Evolution for global search with L-BFGS-B for local refinement, alongside the ROUT method for outlier detection and information criteria (AIC, BIC) for automated phase selection. This modular framework ensures reproducibility and numerical stability, providing researchers in environmental biotechnology a powerful tool to analyze multiphasic substrate degradation dynamics.\ \ Gustavo Mockaitis Open Capsule

Computer Science\ \ 9 | Jan | 2026\ \ SORA ATMAS: Adaptive Trust Management and Multi-LLM Aligned Governance for Future Smart Cities\ \ SORA-ATMAS (Security & Operational Response Agent - Adaptive Trust Management System) is a cutting-edge governance framework for smart cities that integrates:\ \ Dual-Chain Blockchain Architecture: Agentic Blockchain for edge-level provenance + SORA Blockchain for centralized governance\ Multi-LLM Reasoning Ensemble: GPT-4, Grok, and DeepSeek for policy-aligned semantic reasoning\ Three Intelligent Agents: Weather, Traffic, and Safety agents with specialized perception capabilities\ Adaptive Trust Regulation: Context-aware risk assessment with dynamic threshold enforcement\ Real-time Governance: MAE-based LLM selection with error-directed feedback loops.\ \ Usama Antuley et al. Open Capsule

Computer Science\ \ 7 | Jan | 2026\ \ BERTopic-based topic modeling in scientometric analysis of Blockchain integrated Digital Twin Security\ \ This capsule contains the complete Python implementation used to reproduce the topic modeling and scientometric analysis reported in the associated IEEE Transactions on Systems, Man, and Cybernetics: Systems manuscript. The workflow applies an unsupervised BERTopic-based pipeline to Scopus bibliographic metadata to identify thematic structures and research trends in blockchain security.\ \ The capsule includes preprocessing scripts, embedding generation using a transformer-based language model, density-based clustering, and topic representation refinement through maximal marginal relevance. The bibliographic dataset used in the analysis is hosted on IEEE DataPort and is referenced within the capsule to ensure reproducibility without duplicating licensed data. The provided code enables transparent replication and extension of the reported results.\ \ KM Charul Open Capsule

Engineering\ \ 6 | Jan | 2026\ \ Embedded Digital Twin for Power Semiconductors with Adaptive Learning and Parameter Prediction\ \ This code accompanies the reference paper titled “Embedded Digital Twin for Power Semiconductors with Adaptive Learning and Parameter Prediction.”\ Place the files Embedded_DigitalTwin_with_AdaptiveLearning.m and Data.m in the same folder and run the main script. The hyperparameters should be adjusted according to the reference paper.\ \ Alireza Mehrabi Open Capsule

Physics\ \ 6 | Jan | 2026\ \ RBG-Maxwell: A GPU-Accelerated Software Package for Simulation of Collisional Plasmas with Open Boundaries\ \ RBG-Maxwell is an open-source, GPU-accelerated software package for solving the coupled relativistic Boltzmann-Maxwell equations, specifically designed for simulating collisional plasmas with open boundaries. The software uniquely integrates a grid-based Boltzmann equation solver with an electromagnetic field solver based on Jefimenko's equations, which naturally provide absorbing boundaries without reflections. Implemented in Python and optimized for multi-GPU clusters using CuPy, Numba, and Ray, RBG-Maxwell enables large-scale, high-performance plasma simulations. The package fills a critical gap in available tools for fully kinetic, open-boundary plasma modeling, with applications ranging from quark-gluon plasma in heavy-ion collisions to high-altitude nuclear explosion events.\ \ Jian-Nan Chen, Jun-Jie Zhang, Ming-Yan Sun & Yong-Dong Li Open Capsule

Earth Sciences\ \ 9 | Jan | 2026\ \ European forest carbon and biodiversity policies have a limited win-win potential\ \ Data associated to the study "European forest carbon and biodiversity policies have a limited win-win potential"\ \ Lorenzo Balducci Open Capsule

published in Nature Communications

Computer Science\ \ 5 | Jan | 2026\ \ Fixed-Parameter Algorithms for the Tree Containment Problem on Multifurcating Phylogenetic Network\ \ In recent years, phylogenetic trees have proved insufficient for representing certain evolutionary phenomena, particularly reticulation events commonly observed in nature. To overcome this limitation, phylogenetic networks were introduced as a richer model capable of capturing such events. The interplay between phylogenetic trees and networks naturally gives rise to the Tree Containment Problem, a fundamental computational challenge. This problem is NP-hard even for binary networks and has therefore attracted substantial research interest. Earlier work has mainly focused on exponential-time algorithms for the binary setting or on algorithms tailored to specific biologically motivated classes of networks. Prior to this paper, the best-known runtime for fixed-parameter algorithms in the binary case was O(1.618kn2)O(1.618^k n^2)O(1.618kn2), where kkk and nnn represent the reticulation number and the number of vertices in the phylogenetic network, respectively. In this paper, we focus on the Tree Containment Problem on rooted multifurcating phylogenetic networks. Firstly, we study the problem parameterized by the reticulation number kkk of the network and propose a parameterized algorithm with a runtime of O(1.618km3)O(1.618^k m^3)O(1.618km3), where mmm is the number of arcs in the phylogenetic network. Afterwards, we adapt the algorithm to solve the problem parameterized by the level number lll, with a runtime of O(1.618lm3)O(1.618^l m^3)O(1.618lm3). Since l≤kl \le kl≤k for any phylogenetic network, the level-parameterized algorithm is asymptotically more efficient when lll is significantly smaller than kkk.\ \ Jingyi Liu, Zhanglian Lin, Xin Zeng & Feng Shi Open Capsule

Engineering\ \ 5 | Jan | 2026\ \ A constitutive framework for non-isothermal plasticity through micro-mechanism informed artificial neural networks\ \ Physics-informed neural network framework for constitutive modeling of aluminum alloys during non-isothermal hot forming. Combines metallurgical evolution equations (Kocks-Mecking dislocation dynamics, Hall-Petch grain refinement, JMAK precipitation kinetics) with deep learning through a dual-pathway architecture. Enforces thermodynamic consistency via penalty terms. Requires experimental data (500+ points) covering target process conditions. PyTorch implementation with MIT license.\ \ Yo-Lun Yang Open Capsule

Associated article published in Engineering Applications of Artificial Intelligence

Physics\ \ 3 | Jan | 2026\ \ Frozen atom analysis in "Cooperative atomic motion during shear deformation in metallic glass"

[Frozen atom analysis \ This page has been prepared for the peer review of the manuscript titled "Cooperative Atomic Motion During Shear Deformation in Metallic Glass" submitted to Nature Communications.\ It provides access to the short Python script implementing the core algorithm of the manuscript— frozen atom analysis—allowing reviewers or readers to examine the algorithm and perform test runs as part of the review process.\ Contents\ \

• code/frozen_atom_analysis.py Python script to run the frozen atom analysis\
• data/1e09_002/xy_m/m445 Subdirectories including LAMMPS dump files to reproduce the histogram in figure 1c of the manuscript\
• results/displacement_histogram.png Histogram shown in Figure 1c of the manuscript\
• results/displacement_norms.csv CSV file including D_f parameter for each atom\ \ Usage](/content/explore/4e582ef1-bd41-4dfe-81db-778c4587cee3?page=2&filter=all/index.html)

Reproducible run - Check the figure in results/displacement_histogram.png. This should be the same as Figure 1c in the manuscript ( https://arxiv.org/pdf/2503.14903). The atomic group with small DfD_\text{f}Df​ is the STZ core detected in the frozen atom analysis.

To prepare input files

This code requires LAMMPS dump files that record atomic configurations before and after relaxation, in order to calculate the DfD_\text{f}Df​ parameter in the manuscript. Since preparing a complete input file set can be time-consuming, an example dataset is provided in the directory data/1e09_002/xy_m/m445 for convenience.

The procedure for generating these input files is outlined below (for further details, please refer to the manuscript):

1. Perform an AQS (Athermal Quasi-Static) shear deformation simulation using LAMMPS. Examine the resulting stress–strain curve, and extract atomic configurations just before each stress drop event. In this example, we use a particular event labeled m445.
2. For each atom in the configuration, create a separate directory named after its atom ID. See data/1e09_002/xy_m/m445 for an example, and proceed to the next step.
3. Using the configuration from step 1, perform a LAMMPS relaxation calculation in each directory. In each calculation, the atom corresponding to the directory name (i.e., the atom ID) is frozen.

A sample LAMMPS input file for this purpose is provided as misc/frozen.inp. 4. After completing the above preparation for each event, run frozen_atom_analysis.py. This script analyzes the resulting atomic displacements and computes the $D_\text{f} parameter.

Reference and citing

https://arxiv.org/pdf/2503.14903

Author

Yoshinori Shiihara, Toyota Technological Institute

Contact

shiihara[at]toyota-ti.ac.jp

Yoshinori Shiihara et al.

Open Capsule

published in Nature Communications

[Bioinformatics\ \ 3 | Jan | 2026\ \ Biologically-informed integration of drug representations for breast cancer treatment using deep learning\ \ Project introduction \ \ Approximately 50% of breast cancer patients receiving neoadjuvant therapy do not achieve pathological complete response (pCR). Accurately predicting treatment response and selecting the optimal therapeutic strategy are unmet major challenges. To build a model with such functions, learning the interactions between tumor and treatments is necessary. Previous efforts to build response predictors have not incorporated this knowledge. Here, on the basis of data from 4,371 breast cancer patients undergoing neoadjuvant therapy across 31 datasets with transcriptomic and treatment information, we developed GDnet, an interpretable deep learning model integrating drug representations and tumor transcriptome profiles—functioning as a digital organoid—to predict the response to neoadjuvant therapy in breast cancer patients and aid in the selection of optimal treatment strategies. We demonstrate that GDnet (area under the curve (AUC) = 0.725) significantly outperforms transcriptome-only model (AUC = 0.683) in predicting treatment response. Then we conducted two series of in-silico simulated clinical trials based on I-SPY2 trial and external validation datasets respectively and showed that GDnet can optimize treatment selection and significantly increase the pCR rate in all trials. Moreover, the odds ratios (OR) for pCR increase from 1.6 to 2.5 linearly as optimization intensifies. Overall, biologically informed deep learning integrating drug representations and biological omics profiles can function as a digital organoid to optimize breast cancer neoadjuvant treatment selection and may have broader applications across various treatment settings and cancer types.\ \ Code description\ \

• "main.py" file:\
  • This file mainly contain the codes to train GDnet in "3_model_construction_and_training.ipynb". Note: training of Gnetens and GNNetens are not performed give their high demand of cumputing resource.\
  • Because the environment seems do not contain R packages, codes written in R are not included in the main script.\
• Other "*.ipynb" files:\
  • Codes are written in jupyter notebooks with R and python kernel separately. \
  • The notebook files are ordered by the prefixal number (start from 1, end with 9).\ \ Data description\ \
• "Cleaned_data": the patient transcriptome and clinical information\
• "GDSC_and_CTRP_data": GDSC and CTRP transcriptome and drug sensitivity data\
• "Reactome": original data downloaded from Reactome website\
• "Masks_with_prior_knowledge": generated masks containing the prior knowledge in Reactome\
• "Model_performance": the model performance saved for "4_model_performance_comparason.ipynb"\
• Trained_models: the trained models and the model predictions saved for "4_model_performance_comparason.ipynb" and "7_trial_group_aggignment_and_pCR_calculation_and_regimen_interpretation.ipynb"\ \ To train the model\ \
• run "main.py" script\
• trained models and model performance will be saved in "results"\ \ If you have any other questions, please feel free to let us know.\ \ Hewei Ge](/content/explore/b76d702f-e144-4149-ba36-377d20b20525?page=2&filter=all/index.html) Open Capsule

published in Nature Communications

Engineering\ \ 1 | Jan | 2026\ \ Machine Learning–Driven Design of Engineered Cilia Enables Hybrid Operations in Acoustic Microrobots\ \ The code include the ML model to predict cilia resonance peak features and compared to FEA results.\ \ Yun Ling et al. Open Capsule

published in Nature Communications

Biology\ \ 24 | Feb | 2026\ \ Selective control of prefrontal neural timescales by parietal cortex\ \ Intrinsic neural timescales quantify how long spontaneous neuronal activity patterns persist, reflecting dynamics of endogenous fluctuations. We measured intrinsic timescales of frontal eye field (FEF) neurons in rhesus macaques and examined their changes during posterior parietal cortex (PPC) inactivation. We observed two distinct classes of FEF neurons based on their intrinsic timescales: short-timescale neurons (25 ms) and long-timescale neurons (100 ms). Short-timescale neurons showed stronger transient visual responses, suggesting their role in rapid visual processing. In contrast, long-timescale neurons exhibited stronger sustained salience representation, suggesting a role in spatiotemporal integration for maintaining stimulus-driven attention. During PPC inactivation, intrinsic timescales increased in both neuron types, with a significantly larger effect in short-timescale neurons. In addition, PPC inactivation disrupted salience computation, particularly in long-timescale neurons. Here, our findings provide the first causal evidence linking intrinsic local neural timescales to long-range inter-area communications and show the presence of at least two distinct network motifs that support different neuronal dynamics and functional computations within the FEF.\ \ Orhan Soyuhos, Marc Zirnsak, Rishidev Chaudhuri & Xiaomo Chen Open Capsule

published in Nature Communications

Medical Sciences\ \ 31 | Dec | 2025\ \ The efficacy and safety of tislelizumab plus anlotinib in advanced pulmonary sarcomatoid carcinoma\ \ The efficacy and safety of tislelizumab plus anlotinib as first-line treatment in advanced pulmonary sarcomatoid carcinoma: a single-arm phase II trial\ \ Zhimin Zeng, Wenqian Huang & Anwen Liu Open Capsule

Engineering\ \ 30 | Dec | 2025\ \ LCSS_A switched adaptive control approach to reduce sensing needs in trajectory tracking problems

https://ieeexplore.ieee.org/document/11283038

This letter considers an autonomous agent that intermittently acquires state measurements to maintain trajectory tracking performance. The objective is to minimize sensing needs by extending periods of sensor-denied operation. A Lyapunov-based adaptive switched systems approach is developed, where the agent uses the intermittently acquired state measurements to learn the system model. The learned system models are then used during sensor-denied intervals to extend their length while maintaining tracking performance. The design uses a modeling error-dependent bound on the duration of the sensor-denied intervals to progressively reduce sensing needs as the modeling error decreases. The effectiveness of the developed technique is verified in a simulation study.

Muzaffar Qureshi et al.

Open Capsule

Associated article published in IEEE Control Systems Letters

[Mathematics\ \ 30 | Dec | 2025\ \ Cross-Sectional 3D Surface Reconstruction: Benchmarking Python, Numba, and PyTorch Vectorization on CPU/GPU\ \ This capsule provides a fully reproducible Python pipeline for 3D geometric reconstruction, implemented using multiple computational backends and execution models. The code demonstrates how the same reconstruction problem can be solved using pure Python, NumPy + Numba (CPU), Numba with CUDA, PyTorch on CPU, and PyTorch on GPU, enabling direct comparison of computational interfaces and hardware acceleration strategies.\ \ The capsule is designed as a standalone, non-interactive execution, compatible with Code Ocean’s reproducible run framework. All scripts are executed sequentially from a single entry point, and no command-line user input is required.\ \ Reconstruction outputs are visualized using Matplotlib in non-interactive (Agg) mode, and all figures are collected into a single multi-page PDF file. Each page contains a clearly labeled caption describing the computational backend, execution mode (CPU or GPU), and interface used to generate the corresponding reconstruction.\ \ This capsule is intended for:\ \

• Demonstrating reproducible 3D reconstruction workflows\ \
• Comparing Python-based computational backends\ \
• Showcasing GPU acceleration using Numba and PyTorch\ \
• Supporting research transparency and long-term archival\ \ \ The code is modular, extensible, and suitable for execution on both CPU-only and GPU-enabled environments.\ \ Muhammad Zia Afzal, Muhammad Murtaza Yousaf & Shahid Saeed Siddiqi](/content/explore/7e88a694-c002-4a2a-98f0-a7ea32467e29?page=2&filter=all/index.html) Open Capsule

Computer Science\ \ 30 | Dec | 2025\ \ Evaluating ML and Hybrid Models for Social Media Bot Detection\ \ This compute capsule provides reproducible code and experiments supporting the manuscript “Evaluating the Efficacy of Machine Learning Models in Detecting Social Media Bots: An Analysis Using Publicly Available Datasets.” It includes data preprocessing, machine learning and deep learning models, a hybrid CatBoost–DNN ensemble, and evaluation using accuracy, F1-score, ROC-AUC, and SHAP-based interpretability on the VKontakte Users vs Bots dataset.\ \ Armin Khoshkar Open Capsule

Engineering\ \ 30 | Dec | 2025\ \ Explainable mechanism for production process anomalies based on digital twin\ \ In the manufacturing sector, abnormal production (AP) can disrupt production schedules, leading to significant economic and reputational losses for manufacturers. To address this issue, in this study, we present an explainable mechanism for production process anomalies (EM2PA) designed to clarify the complex coupling relationships among various manufacturing factors, analyze the impact of these factors on the production process, identify AP, provide explanations for its causes, and enable trace-back analysis. EM2PA consists of three modules: the data augmenter (DAr), the influence factor recognizer (IFRr), and the causal interpreter (CIr). Specifically, the DAr generates small sample data of AP, the IFRr decouples the complex coupling relationships and identifies the factors influencing AP, and the CIr provides causal explanations. Furthermore, through a case study based on the actual production process of a discrete manufacturing workshop, we demonstrate the effectiveness of EM2PA in identifying the root causes of problems, while highlighting the importance of explainability and causal analysis of production process anomalies.\ \ Weiwei Qian et al. Open Capsule

published in Nature Communications

Bioinformatics\ \ 29 | Dec | 2025\ \ CARMSeD: Multi-Task Learning for Synergy Class and Signal Prediction\ \ This capsule implements the CARMSeD model pipeline described in the manuscript "AI-Guided CAR Designs and AKT3 Degradation Synergize to Enhance Bispecific and Trispecific CAR T Cell Persistence and Overcome Antigen Escape." It includes scripts for functional signal labeling, synergy class assignment (CARMSeD), multi-task model training, and inference on new CAR sequences. Sample data and trained model outputs are provided to enable full reproducibility.\ \ Nisha Chaudhary Open Capsule

published in Nature Communications

Engineering\ \ 27 | Dec | 2025\ \ Create figures for Fourier Series Characteristic Function method of Nonlinearity Analysis of Multi-Carrier Signals\ \ Abstract:\ Nonlinearities in power amplifiers adversely affect multi-carrier modulation techniques. Accurate prediction of nonlinear distortion is essential for making design trade-offs between output power and network throughput. We use the series form of the characteristic function (ch.f.) method to predict distortion spectra for sparse multi-carrier transmissions. This method results in efficient calculations of individual signal and distortion components. The method is validated both theoretically and practically. Theoretical validation is performed by modeling the signal as a bandpass Gaussian process that is hard limited, and it is shown that the series ch.f. method produces results that are identical with the classical Price’s theorem. Practical validation is shown by considering an orthogonal frequency division multiplexing (OFDM) signal with a fragmented spectrum which is then applied to an amplifier driven into compression for which application of Price’s theorem is difficult, and the predicted output spectrum corroborates laboratory measurements. Part of the computational efficiency is realized in that the nonlinearity can be expressed as the fast Fourier transform (FFT) of samples of its forward scattering parameter (i.e., S21) or transconductance function (including AM-PM effects), and distortion contributions of the signal can be expressed as numerical autoconvolutions of the clean spectrum. Signal-to-distortion ratio (SDR) can be easily computed and parameterized across variables of interest, such as overdrive level.\ \ Cameron Pike Open Capsule

Associated article published in IEEE Transactions on Circuits and Systems II: Express Briefs

Physics\ \ 27 | Dec | 2025\ \ Multiphysics model for laser-driven particle acceleration

The model is solved for 5 different lasers (TITAN, TPW, OMEGA EP, ORION and L4f ATON).\ \ For the exhibition of the code in Code Ocean, I print in the outputs of some fundamental variables:\ -Cold preplasma Scale Length\ -Hot preplasma scale length\ -Cold preplasma electron temperature\ -Number of free electrons per atom in preplasma\ -Peak relativistic electron temperature\ -Maximum relativistic electron temperature\ -Peak positron energy\ -Positron yield\ -Reflux coefficient\ -Peak relativistic electron projected length\ -Maximum energy relativistic electron projected length\ -Photon average path length\ -Sheath field energy\ -Laser power threshold for nonlinear Breit-Wheeler initiation\ -Laser power threshold for nonlinear Breit-Wheeler to become the governing pair production mechanism \ \ However, user can print as output anyone of the variables included in this code, by reducing the respective semicolon and organizing the results in tables as presented in this code for the aforementioned outputs.\ \ Inputs of this code are the laser parameters, presented in all_data.txt. The numerical values from this .txt file, were inserted manually in the code in the "alldata" matrix.

When using this model, please cite:\ Alexopoulou, V. E. Advanced modeling of short and ultrashort laser irradiation of metals in micro-nano down to sub-atomic scale. PhD thesis - National Technical University of Athens (2025). http://dx.doi.org/10.26240/heal.ntua.30238 & Alexopoulou, V. E., Markopoulos, A. P. Highly-accurate, fully-coupled heat transfer-hydrodynamic-electromagnetic simulation for modeling and optimizing laser-driven particle acceleration for laboratory astrophysics. ResearchSquare (2025). https://doi.org/10.21203/rs.3.rs-7562653/v1

Vasiliki E. Alexopoulou

Open Capsule

Computer Science\ \ 27 | Dec | 2025\ \ Reasoning-Stable Fuzzy Equilibrium Learning for Severity-Aware Clinical Decision Intelligence in CKD under Incomplete Evidence\ \ This capsule provides a reproducible implementation of the baseline pipeline and reasoning-stability evaluation workflow used in the manuscript.\ The code loads anonymized clinical records, performs preprocessing, trains reference models (Logistic Regression, XGBoost, MLP, and TabNet), and evaluates prediction consistency under simulated missing-evidence conditions.\ \ The capsule generates:\ • model performance summaries\ • reasoning-drift stability metrics\ • a demonstration plot illustrating stability behaviour\ \ The implementation is provided for transparency and reproducibility. It does not include full ablation studies, extended analyses, or datasets beyond what is necessary to reproduce the core experimental pipeline.\ \ Deblina Kar et al. Open Capsule

Engineering\ \ 15 | Mar | 2026\ \ Neural Value Alignment: human-AI collaboration under goal-action ambiguity\ \ This is the code used to replicate the results in the manuscript titled: Neural Value Alignment: human-AI collaboration under goal-action ambiguity.\ \ Xin Xu et al. Open Capsule

Biology\ \ 23 | Dec | 2025\ \ Comprehensive identification of cis-regulatory enhancers and promoters provides insights into tissue-specific gene regulation in the sheep reference genome\ \ We systematically characterized 274,682 enhancers and 25,975 promoters across 24 tissues in sheep using six high-resolution assays: ChIP-seq, ATAC-seq, CAGE-seq, RRBS, WGBS, and RNA-seq. The study provides a robust framework for exploring cis-regulatory mechanisms and tissue-specific regulation, advancing the functional annotation of the sheep reference genome. This capsule provides the analysis code for this study.\ \ Shangqian Xie Open Capsule

published in Nature Communications

Social Sciences\ \ 20 | Dec | 2025\ \ From Faces to Politics: Vision-Language Models (Sometimes) Link Visual Demographic Characteristics to Ideological Labels\ \ This capsule contains the replication materials (data and code) for the paper 'From Faces to Politics: Vision-Language Models (Sometimes) Link Visual Demographic Characteristics\ \ to Ideological Labels'. It includes scripts to reproduce the data cleaning, statistical modeling, and generation of all tables and figures regarding ideological bias in Vision-Language Models (GPT-4o, GPT-4o-mini, LLaVA) versus human survey benchmarks.\ \ Soyeon Jeon, Messi H.J. Lee, Jacob M. Montgomery & Calvin K. Lai Open Capsule

Engineering\ \ 20 | Dec | 2025\ \ Matlab Code - An algorithm for calculating the Final Yield and Capture Losses of a photovoltaic system at the string, MPPT and inverter levels\ \ The code processes and analyzes operational data from a photovoltaic plant to evaluate its monthly energy yield at different levels of aggregation (string, MPPT and inverter). Initially, it loads and cleans the experimental data, managing missing values and interpolating irradiance when necessary. It then calculates standardized indicators widely used in the literature, such as reference yield (Yr), array yield (Ya), capture losses (Lc), and performance ratio (PR). The results are aggregated hierarchically from the strings to the inverter. Finally, the script generates monthly graphical representations that allow for visual comparison of yields and losses.\ \ Francisco José Muñoz-Rodríguez et al. Open Capsule

Engineering\ \ 20 | Dec | 2025\ \ Dynamic mode decomposition of higher order systems\ \ Joel A. Rosenfeld, Benjamin Russo, and Rushikesh Kamalapurkar, Dynamic mode decomposition of higher order systems, Automatica, to appear, 2026.\ \ Joel A. Rosenfeld, Ben Russo & Rushikesh Kamalapurkar Open Capsule