deep learning

Single-cell foundation models transfer across deep evolutionary distances

This paper evaluates the transferability of single-cell foundation models across deep evolutionary distances using developmental atlases of Dictyostelium discoideum and Caenorhabditis elegans.

Neanderthal introgressed ancestry reveals human genomic regions enriched with recessive deleterious mutations

The paper uses Neanderthal ancestry patterns to identify genomic regions enriched with recessive deleterious mutations using a machine learning classifier.

ChAI-Net learns a regulatory code for chromatin co-accessibility

The paper introduces ChAI-Net, a deep learning model that predicts chromatin co-accessibility from DNA sequences in mouse embryonic stem cells, revealing sequence motifs that govern coordinated TF binding.

AI-driven discovery and engineering of human endogenous nanocage proteins for mRNA delivery

The paper develops DeepDelivery, an AI platform that identifies human proteins for mRNA delivery, validating 80% of top candidates for efficient, non-inflammatory delivery.

Digital pathological classification of peripheral neuroblastic tumors across multi-continent cohorts

The paper develops NB-PathAssist, a computational framework for automated classification of peripheral neuroblastic tumors using whole-slide histology images, achieving high accuracy in prognostic grouping.

Bioactivity foundation model for hit identification and hit-to-lead optimization from sequence data

PSICHIC+ is a bioactivity foundation model that predicts binding affinities from sequence data, improving hit identification and hit-to-lead optimization using heterogeneous bioactivity data.

SMFnet reveals principles of combinatorial transcription factor assembly

The paper introduces SMFNet and PRISM to analyze how transcription factor assemblies influence chromatin accessibility and identifies sequence-encoded principles of combinatorial TF assembly.

DCAF: Unifying Data Validity, Evaluation Validity and Predictive Modeling for Single-cell Perturbation Simulation

DCAF improves single-cell perturbation simulation by addressing data, evaluation, and model validity, outperforming baselines in generalization and recapitulating regulatory heterogeneity in mouse brain.

Extending digital biology: bacterial survival and morphological heterogeneity under antibiotic stress

This study develops a workflow to generate morphological signatures of bacterial progeny under antibiotic stress, revealing distinct morphologies linked to survival and antibiotic type.

Multi-scale haplotype signatures distinguish among evolutionary processes

SHADES, a deep learning framework, distinguishes evolutionary processes like neutrality, selection, and introgression using haplotype data from humans, outperforming existing methods.

COMBO-RATE: An experimentally validated bioinformatic tool to identify promiscuous HLA restrictions

COMBO-RATE identifies promiscuous HLA restrictions by analyzing immune response data, outperforming existing tools and revealing broader HLA allele coverage for T cell epitopes.

Fox: Predicting rates of host-symbiont co-evolution events with association-aware attention

Fox predicts host-symbiont co-evolution event rates directly from sequence alignments without tree reconstruction, achieving high accuracy and speed compared to existing methods.

Protein co-scientist aided simulations of the structural and functional impact of O-GlcNAc Transferase TPR domain deletions and amino acid substitutions associated with X-Linked Intellectual Disability

The paper uses computational methods to model structural and functional impacts of TPR domain deletions and amino acid substitutions in OGT linked to X-Linked Intellectual Disability.

NeuroPack: Advanced Learnable Packaging Toolkit Towards Ultra-High Compression Ratio for Volumetric Neuroimaging

NeuroPack compresses 3D neuroimaging data with high fidelity and ultra-high compression ratios using a hyperprior-driven deep learning toolkit.

Network-guided identification of candidate biomarkers for COPD-associated comorbidities using deep learning and real-world transcriptomic data

The paper develops a network-guided deep learning approach to identify biomarkers for COPD-associated comorbidities using transcriptomic data and validates them in public datasets.

Kontext: Kernel-based method bridging spatial domain identification and cell-cell communication

Kontext identifies spatial domains and cell-cell communication by integrating ligand-receptor interactions, outperforming existing methods in spatial transcriptomics.

Deep Learning Based Cell Segmentation of 3D Confocal Images of Wood for Enzymatic Hydrolysis Applications

This paper develops a 3D U-net model for segmenting wood cells in confocal images to improve predictions of enzymatic hydrolysis efficiency.

Spatial Single-Cell Mapping of Transcriptional Differences Across Genetic Backgrounds in Mouse Brains

The paper introduces ATLAS, a method to map transcriptional differences across genetic backgrounds in mouse brains, revealing significant cell type distribution changes between sexes and genetic models.

VoroTIDE: Protein domain segmentation through independent hierarchical searches using Voronoi tessellation and physical self-containment

VoroTIDE segments protein domains using Voronoi contact areas and self-containment scores, achieving strong performance on benchmark datasets.

Mechanistic gene-regulation-based trajectory inference with bowside

The paper introduces bowside, a neural model for trajectory inference and gene regulatory network reconstruction in single-cell RNA sequencing data, showing it recovers known regulatory motifs in hematopoiesis and reprogramming.

IDPForge: Deep Learning of Proteins with Global and Local Regions of Disorder

IDPForge generates all-atom ensembles of intrinsically disordered proteins and regions using a transformer diffusion model, showing good agreement with experimental data.

Deep Mutational Scanning of the GPCR Rhodopsin Reveals Pleiotropic Mutational Effects and Their Roles in Modulating Light Sensitivity

The paper uses deep mutational scanning to study pleiotropic effects in rhodopsin, revealing how mutations impact multiple molecular phenotypes and shape light sensitivity.

A comprehensive map of the sequence basis of transcription initiation in humans

The paper develops deep learning models to predict transcription initiation patterns from DNA sequences across human tissues and cancers.

Advancing Spatial Pattern Integration in Species Distribution Modelling With Computer Vision: How Grain Size and Extent Concepts Shape Predictions of Emerging Aquatic Insects

This paper evaluates how grain size and window extent affect Deep SDMs for predicting aquatic insect distributions, finding significant performance variations with different resolutions and extents.

ARG-PASS v2.0 expands global resistome surveillance to reveal previously unrecognized antibiotic resistance genes reaching the human gut

ARG-PASS v2.0 identifies previously unrecognized antibiotic resistance genes in environmental samples and links them to human gut microbiomes through dietary exposure.

Aβ42 facilitates the activation of Discoidin Domain Receptor 2 and Its nuclear enrichment in Alzheimer's Disease Model

This study shows Aβ42 activates DDR2 in Alzheimer's disease models by non-canonically bypassing collagen ligands and promotes its nuclear enrichment, with structural insights into the mechanism and potential therapeutic inhibition.

Agentic cryo-EM structure determination with CryoPilot

CryoPilot uses two language model agents to automate cryo-EM structure determination, improving resolution and model accuracy through adaptive decision-making.

Digital twins forecast individual immune trajectories

The paper introduces ImmuneTwins, a digital twin framework that personalizes immune response models for individuals, improving prediction accuracy of immune trajectory changes.

Abundance-weighted analysis and deep functional metagenomic selection reveals unrecognized diversity of soil-derived streptothricin acetyltransferases

The paper uses deep functional metagenomics to discover new streptothricin acetyltransferase genes in soil, revealing previously unknown resistance mechanisms.

Does a map-level insulation constraint improve sequence-based prediction of 3D genome folding? A Stage 1 Registered Report with a reverse-complement-equivariant state-space model (HiFold-Mamba)

The paper evaluates whether insulation-based losses improve sequence-based prediction of 3D genome folding using a reverse-complement-equivariant state-space model, finding improvements in boundary and loop prediction.