deep learning

Label-free identification of bacterial species using adapted vision transformers

This paper develops a vision transformer model for rapid bacterial species identification from time-lapse images without fluorescent labels, achieving high accuracy on clinical isolates.

Winnow-tax: sensitive and precise taxonomic profiling of low-coverage organisms in shotgun metagenomes

Winnow-tax improves taxonomic profiling of low-coverage shotgun metagenomes by combining sensitive read-level classification with genome-level confirmation, achieving better precision-sensitivity balance than existing methods.

Multilayer analysis reveals extensive hidden population complexity in a dairy starter microbiome comprising few species

The paper analyzes a dairy starter microbiome with low species richness using genomic and metagenomic data to reveal hidden population complexity and phage resilience.

Predicted structure of the 89-kDa invasion-tip protein of Holospora obtusa, an endonuclear symbiont of Paramecium caudatum, suggests a pH-dependent conformational change

The paper predicts the structure of a 89-kDa invasion-tip protein from Holospora obtusa and suggests a pH-dependent conformational change may aid in escaping the host digestive vacuole.

EventHorizon: A Foundation Model for Clinical Flow Cytometry

EventHorizon is a foundation model for clinical flow cytometry that uses self-supervised learning to generate embeddings for diagnostic use, achieving high accuracy in identifying genetic abnormalities.

Pheno-MYCN maps the morphological footprint of MYCN amplification in paediatric neuroblastoma

Pheno-MYCN links slide-level MYCN prediction to interpretable morphological sub-populations in neuroblastoma, revealing a continuous gradient of MYCN-associated features.

Vision Normalizing Flows for the probability-informed detection of banana diseases from in-field images

The paper develops a vision normalizing flow model to detect banana diseases from in-field images with high accuracy, using pre-trained embeddings and probabilistic density estimation.

JEPA-SORT: World-Model Embeddings Improve Deep-Sea Multi-Object Tracking on Visually Homogeneous Fauna

JEPA-SORT improves deep-sea multi-object tracking by using world-model embeddings to maintain stable identities on visually homogeneous fauna, outperforming standard MOT algorithms in IDF1, MOTA, and BDU.

Probing Genomic Foundation Models with Splice-Variant Perturbations

This paper evaluates genomic foundation models using splice-variant perturbations, finding they match task-specific networks and generalize across mutation types.

Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery

This paper introduces a generative AI framework for drug discovery that efficiently navigates large chemical spaces while ensuring synthetic feasibility and functional success.

An AI-Enabled Translational Drug Discovery Framework for NALCN Channelopathy: From High-Throughput Screening to Therapeutic Candidate

This paper develops an AI-driven drug discovery framework for NALCN channelopathy, identifying aprepitant as a lead therapeutic candidate through HTS, structural modeling, and experimental validation.

Combining AlphaFold 3 with microscale thermophoresis for aptamer optimisation

The paper combines AlphaFold 3 with microscale thermophoresis to study aptamer-protein interactions, finding that structural predictions are not reliable predictors of binding strength.

Predicting viral host range for reverse zoonosis surveillance

The paper predicts viral host range using a combination of experimental screening and deep learning to identify potential animal reservoirs for reverse zoonotic transmission.

DEPICT enables sample-centric therapeutic prioritization across evolving and spatially heterogeneous tumor states

DEPICT prioritizes therapies within tumor contexts using deep learning, showing effectiveness across different tumor states and spatial heterogeneity.

Genome reorganisation and expansion shape 3D genome architecture and define a distinct regulatory landscape in coleoid cephalopods

The paper investigates how genomic changes in coleoid cephalopods shape 3D genome architecture and regulatory landscapes, revealing species-specific chromatin loops and their role in neural development.

Inferring seagrass meadow resilience from self-organized spatial patterns

The paper uses synthetic data generated from a mechanistic model to train deep learning networks that classify seagrass meadow resilience from spatial patterns, showing effective generalization to real-world data.

De novo protein design in an expanded chemical space

This paper introduces CheMoDesign, a framework for de novo protein design with user-specified chemistry, demonstrating its ability to generate proteins with chemistry-dependent recognition and reactivity.

Fine-tuning genomic models on rare splicing events identifies a novel biomarker of TDP-43 pathology

Frustration Quenching and Network Topology of the Energy Landscape as Primary Determinants of Protein-Ligand Binding Pose Prediction by Deep Learning Models

This paper shows that deep learning models' performance in predicting protein-ligand binding poses depends on the energy landscape's frustration quenching and network topology, not algorithmic flaws.

MINT: Predicting Enzyme-Cofactor Specificity by Multimodal Knowledge Transfer from Protein-Ligand Interactions

MINT predicts enzyme-cofactor specificity using a multimodal framework with transfer learning, achieving high accuracy in cofactor binding prediction.

Generative inverse design steers human hematopoietic cell fate in vitro

This paper introduces LabCompass, a generative inverse-modeling framework that optimizes culture conditions to steer human hematopoietic cell fate in vitro, demonstrating its effectiveness through iterative experimental validation.

Comparative Analysis of Relative Ligand Binding Free Energy Simulation Methods: Amber-TI, GROMACS-NETI, OpenMM-FEP, and BLaDE-MSLD

This study compares four ligand binding free energy methods using the same force field and finds they show comparable accuracy with no significant performance differences.

DECONVersation: Single Cell Foundation Model-Derived Embeddings for Robust Cell Type Deconvolution of Bulk RNA-seq

DECONVersation uses single-cell foundation model embeddings to improve bulk RNA-seq deconvolution by reducing batch effects and marker gene sensitivity.

A deep learning-generated hASIC1a miniprotein inhibitor confers neuroprotection in stroke

This paper designs a deep learning-generated hASIC1a inhibitor that reduces infarct volume by over 50% in a mouse stroke model.

VesiScope: A stand-alone tool for the automatic detection, quantification and size analysis of GUVs from phase contrast microscopy images

VesiScope detects and analyzes GUVs from phase contrast images using a YOLOv11 model, enabling accurate and efficient label-free quantification.

SIMORGH: Ensemble-Aware Geometric Deep Learning for Apo-State and Cryptic Ligand Binding Site Prediction

SIMORGH predicts ligand binding sites from apo proteins by integrating conformational ensembles, outperforming existing methods in ligand-free settings and identifying cryptic pockets.

Large language models unlock the ecology of species interactions

This paper uses large language models to extract species interaction data from citizen science comments, expanding ecological data coverage and enabling new ecological research.

Illuminating dark genes and dark metabolites: calibrated functional association across kingdoms with autoencoders and correlation

VaLPAS pairs a bi-directional autoencoder with correlation for omics guilt-by-association with calibrated confidence; in L. starkeyi its autoencoder linked 176 unidentified metabolites to known-function proteins at ≥50% confidence.

FragDockRL: A Reinforcement Learning Method for Fragment-Based Ligand Design via Building Block Assembly and Tethered Docking

FragDockRL uses reinforcement learning to design ligands by assembling building blocks and tethered docking, outperforming random search on three protein targets.

Machine learning classifies subcellular Golgi morphology from imaging flow cytometry during inflammatory activation: an exploratory study

This study uses deep learning to classify Golgi morphology from imaging flow cytometry data to distinguish LPS-treated from control cells in human and mouse immune cells.