deep learning

Deep learning models detect neuromuscular disease and quantify functional impairment from video-derived biomechanics data

This paper develops deep learning models to detect neuromuscular diseases and quantify movement impairment using video-derived biomechanics data from smartphone recordings.

Cell type-resolved connectomic reconstruction of the spinal dorsal horn reveals sensory neuron synaptic organization and modality-specific inhibition

The paper presents tools to reconstruct spinal cord circuits and identifies modality-specific inhibitory synapses between sensory neurons and inhibitory neurons.

Accurate ab initio gene prediction in eukaryotes with Tiberius in multiple clades

Extended Tiberius models outscore Helixer and ANNEVO on all 33 genomes from the five clades all three tools cover, and approach BRAKER3 accuracy in four clades without extrinsic evidence.

Beyond RMSD: Deterministic Dlow Matching for Ultrafast and Physically Validated Transition-State Generation

This paper introduces Kairos-GFM, a model that generates transition states with high physical fidelity and low RMSD, outperforming diffusion-based methods but underperforming transport-based ones.

Ultrafast and Accurate Selection of High-Quality Protein Complex Models from Large-Scale Prediction

This paper introduces DeepUMQA-Selection, a framework for efficiently selecting high-quality protein complex models from large candidate pools, showing improved performance and speed.

How the TREX-2 complex associates with the nuclear pore

The paper reports an extended structure of the human nuclear pore complex, revealing TREX-2 components as integral parts of the nuclear ring.

The cysteine-rich domain of SEP15, a selenoprotein co-chaperone of the ER chaperone, UDP-glucose:glycoprotein glucosyltransferase, adopts a novel fold

The paper determines the structure of the cysteine-rich domain of SEP15 using AlphaFold and NMR, revealing a novel helical fold stabilized by disulfide bonds.

Interpreting Convolutional Neural Networks in Population Genetics

This study investigates how convolutional neural networks in population genetics implicitly compute summary statistics and reveals their ability to efficiently approximate long-range linkage disequilibrium.

Genetic risk and inflammatory signaling converge on cell type-specific regulatory programs in type 1 diabetes

This study identifies T1D-associated SNPs that alter chromatin accessibility at TF binding sites in pancreatic islets, linking genetic risk to cell type-specific regulatory programs.

SSAlign: Ultrafast and Sensitive Protein Structure Search at Scale

SSAlign improves protein structure search speed and sensitivity by using language models and a two-stage alignment strategy, outperforming Foldseek on large datasets.

Systematic exploration of predicted quaternary structures within pandemic-relevant viral proteomes

The paper predicts quaternary structures of viral proteins using AlphaFold2 and AlphaFold-Multimer, identifying new insights into vaccinology and protease targets.

pH-sensitive binder design with Proton-PottsMPNN

This paper introduces Proton-PottsMPNN, a model that designs pH-sensitive binders by explicitly modeling protonation states, resulting in pH-dependent PD-L1 binders with transition pH values between 4.0 and 5.8.

HMVA-DDI: A Hierarchical Multi-View Attention Network for Drug-Drug Interaction and Severity Prediction

HMVA-DDI predicts drug-drug interactions and their severity using a two-stage deep learning framework with improved attention mechanisms and task-specific features.

A functionally validated TCR-pMHC database for TCR specificity model development

The paper develops a validated TCR-pMHC database (TCRvdb) by assessing VDJdb entries with a high-throughput platform, finding only 50% of claims are functionally confirmed.

A systematic delineation of 3' UTR regulatory elements and their contextual associations

The paper identifies 3'UTR regulatory elements using synthetic inserts and a deep learning model, revealing their effects on gene expression and nuclear-cytoplasmic partitioning, and highlights disease-relevant variants.

Interpretable models for scRNA-seq data embedding with multi-scale structure preservation

This paper introduces two tools for improving the trustworthiness of single-cell dimensionality reduction by enhancing quality metrics and multi-scale structure preservation.

Hybrid modeling framework for bioprocesses with minimal prior knowledge and limited data

This paper introduces a hybrid modeling framework that combines mechanistic ODEs with neural networks to model bioprocesses using minimal prior knowledge and limited data.

Cancer Evolvability Determines Therapy Outcomes

This paper develops an evolutionary model to show how cancer evolvability affects therapy outcomes, finding that targeted therapy is better for low-evolvability cancers and chemotherapy for high-evolvability ones, with a combined approach effective against facultative evolvability.

This must be the place: deep learning local adaptation

EcoLocator, a deep learning model, predicts geographic and climate origins from genotypes, outperforming traditional methods and showing practical applicability in conservation.

Joint enzyme-reaction retrieval and catalytic optima prediction via multimodal fusion

GERO improves enzyme-reaction retrieval and catalytic optima prediction by integrating multimodal data for condition-aware recommendations.

Evolution and adaptations of the seminal proteome in an insect with traumatic insemination

The paper characterizes the sperm and seminal fluid proteome of bedbugs, revealing evolutionary adaptations linked to traumatic insemination and identifying expanded S-Laps with potential catalytic activity.

Inferring binding specificities of human transcription factors with the wisdom of crowds

The paper benchmarks methods for inferring transcription factor binding specificities using a large community challenge and finds deep learning models outperform traditional methods, though positional weight matrices are competitive.

Deep Lineage: Single-Cell Lineage Tracing and Fate Inference Using Deep Learning

Deep Lineage uses deep learning to analyze time-series single-cell RNA-sequencing with lineage-tracing data to predict cell fate and gene expression profiles more accurately than existing methods.

HipSAFE: automating hip fracture detection on ultrasound imaging using deep learning

HipSAFE uses deep learning to detect hip fractures on ultrasound images, achieving high accuracy comparable to x-rays, potentially improving triage in rural settings.

An explainable machine learning consensus framework for robust estimations of environmental effects on population dynamics

This paper develops a framework to assess explainable ML sensitivity by comparing consensus across model architectures, using synthetic coral data to demonstrate its utility in ecological modeling.

Physics and Morphology Constrained Quantitative Susceptibility Based Segmentation of Cerebral Veins

Dipole-field physics and Frangi vesselness losses lift QSM vein segmentation to centreline Dice 0.70 (U-Net 0.57); derived venous OEF tracks calibrated fMRI better than vessel filtering.

Generative modeling reveals the connection between nuclear morphology and gene expression

COSMIC, a generative model, quantifies the relationship between nuclear morphology and gene expression using paired single-cell data, revealing morphology-associated genes in prostate cancer and spatial transcriptomics of zebrafish embryos.

Deep learning based design of buried hydrogen bond networks with HBDesigner

This paper introduces HBDesigner, a deep learning tool for designing hydrogen-bond networks in proteins, showing improved performance over existing methods.

VRPTR prediction of individual language activation and uncertainty from resting state fMRI

VRPTR predicts individual language activation from resting-state fMRI with improved correlation over baseline models, while providing calibrated uncertainty estimates.

A pathway to next-generation mast cell stabilizers identified through the novel Phytomedical Analytics for Research Optimization at Scale data platform.

This study identifies potential next-generation mast cell stabilizers using a novel data platform and in vitro pharmacology, prioritizing candidates from phytomedicine.