deep learning

A multimodal precision oncology platform integrating genomics,single-cell transcriptomics, and functional drug screening expandstherapeutic options in pediatric brain tumors

Computational design of metalloproteases

This paper designs de novo metalloproteases using RoseTTAFold Diffusion 2, achieving high activity and specificity for cleaving various substrates.

Limits of marker information transfer]{Limits of marker information transfer in cross-panel cytometry imputation

This study evaluates the limits of marker information transfer in cross-panel cytometry imputation, finding that imputation preserves broad lineage classification but not fine-phenotype details, and that patient matching improves cross-platform distribution recovery.

Multistage AI Crop Inspection and Segmentation Framework (MACISF)

MACISF is a deep learning framework for automated fruit quality evaluation, achieving high accuracy in classification, health detection, and defect localization.

REMAG: recovery of eukaryotic genomes from metagenomic data using non-contrastive self-supervised learning

REMAG improves eukaryotic genome recovery from metagenomic data by using non-contrastive self-supervised learning, outperforming existing tools in producing near-complete genomes.

Deep-learning-driven tumor microenvironment profiling improves immunotherapy response prediction

EcoNet improves immunotherapy response prediction by modeling tumor microenvironment signaling networks, outperforming existing methods in AUC and stratifying ccRCC patients by clinical outcome.

Towards Robust Automated Mouse Scratching Behavior Detection for Itch Research

This paper develops Scratch-AID+, an automated mouse scratching detection framework, showing that fine-tuned video foundation models outperform classic models in generalizability and accuracy.

A hippocampal digital Twin reveals latent structure in freely flying bats

This paper develops a functional digital twin of the hippocampus in bats to uncover latent neural organization during natural flight behavior.

BioMod: Biological Module Tokenization for Single-Cell Foundation Modeling

BioMod introduces biological modules as intermediate units for single-cell foundation modeling, achieving state-of-the-art performance in cell-type annotation and drug-response prediction.

A robust and scalable image analysis framework for phenotyping human thoracic aorta

This paper introduces a framework for automated thoracic aorta phenotyping from CT scans, generating anatomical landmarks, meshes, and biomechanical phenotypes.

A multimodal foundation model to integrate histology with spatial molecular programs

SQUALL integrates histology and spatial transcriptomics to enable virtual spatial gene expression profiling and clinical prediction across cancers and platforms.

PanTEon: a cross-kingdom framework to guide the design of transposable element classifiers

PanTEon provides a cross-kingdom framework for reproducible TE classification, benchmarking machine learning methods and highlighting performance variations across TE superfamilies and lineages.

Prospective discovery of LasB inhibitors by integrating generative design, free energy ranking, and experimental validation

This paper discovers three new nanomolar inhibitors of LasB by combining generative design, free energy calculations, and experimental validation.

Pocket-Conditioned Generation and Down-Selection of Non-Oxime Reactivators

This paper compares generative models for designing non-oxime reactivators of paraoxon-inhibited acetylcholinesterase, finding that down-selection methods outperform DrugDiff in practical drug-likeness and synthesis feasibility.

A foundation model for brain organoid network dynamics uncovers intrinsic computation and convergent disease phenotypes

PULSE, a foundation model trained on neural recordings, reveals how disease variants affect network development and how circuits encode information in brain organoids.

VirtualSkin integrates skin-cell identity with intervention-response prediction

VirtualSkin predicts intervention responses in skin cells by integrating transcriptomic data with biological descriptions across multiple datasets.

Beyond the attractor: closed-loop stimulation design for learning multistable neural dynamics

This paper introduces PALDI, a method for active closed-loop stimulation to identify neural dynamics, showing improved performance in noisy and biased observation scenarios.

tRNAZAP: Deep learning toolkit for nanopore tRNA signal analysis

tRNAZAP is a deep learning toolkit for analyzing nanopore tRNA sequencing data, achieving high accuracy in segmentation and classification tasks.

The Human Gut Cell Atlas v1.0: A Unified, Community-Built Reference of Cellular Diversity Across the Gut

The paper presents the Human Gut Cell Atlas v1.0, a unified reference of gut cell diversity, revealing how tissue structure influences cellular composition and enabling disease and organoid analysis.

A Multi-Omic Phenobank Reveals Axes of Glioblastoma Growth, Invasion, and Therapeutic Vulnerability

The paper establishes a multi-omic PDX resource linking GBM invasion phenotypes to molecular programs and therapeutic responses, identifying two invasion modes and a neurodevelopmental factor associated with tumor growth and survival.

Pre-neuronal origins of monoamine and acetylcholine synthesis and transport

This study identifies the evolutionary origins of neurotransmitter synthesis and transport in fungi, showing these functions predate neurons and are linked to lysosomal processes in opisthokonts.

InfluProto: Predicting Cross-Species Transmission of Influenza A Virus via Prototype Learning

InfluProto predicts cross-species transmission of influenza A virus using prototype learning, showing high accuracy in host classification and detecting reassortment effects.

CLARA: Compound Library Active Learning with Reasoning Agents

CLARA uses a language model to make strategy decisions in compound library active learning, outperforming random sampling across multiple protein targets.

Human-specific CCG-repeat insertion in the RBFOX1 promoter implicated in autism and speech enhances EGR1-responsive activity

A human-specific CCG insertion in the RBFOX1 promoter boosts EGR1-driven reporter activity in HEK293T cells, as does an autism-linked CCG-gaining duplication in PTCHD1.

Large scale prospective evaluation of co-folding across hundreds of Mac1-ligand complexes and three virtual screens

On 489 evaluable new Mac1 complexes, AlphaFold3, Chai-1 and Boltz-2 placed over half of ligands within 2 Å, but on docking hit-lists AF3 and Boltz-2 added little over docking.

Inherent limitations of zero-shot mutation prediction in protein engineering

This study evaluates zero-shot mutation prediction models in protein engineering, finding they struggle to identify optimal mutations and predict new functions due to limitations in capturing evolutionary constraints.

Physics-informed neural networks for subject-specific recovery of travelling-wave structure in resting-state fMRI

This paper introduces a physics-informed neural network to recover subject-specific travelling-wave structure in resting-state fMRI by incorporating anatomical priors.

Alignment-Aware Genomic Coordinate Encoding for Long-Range DNA Modeling

This paper introduces USM-AA, an alignment-aware genomic foundation model that improves long-range DNA modeling by incorporating reference genome coordinates and reverse-complement masked learning.

AlphaFold3 and Intrinsically Disordered Proteins: Reliable Monomer Prediction, Unpredictable Multimer Performance

AlphaFold3 pLDDT matches AlphaFold2 at calling disordered residues (Matthews correlation 0.693 vs 0.701); on disorder-containing complexes mean DockQ is similar, but interface features explain less variance (42% vs 58%).

Representing biological concepts as directions in foundation model embeddings

This paper shows that biological concepts can be represented as directions in foundation model embeddings, enabling interpretable and reusable analysis of diverse biological properties.