deep learning

Deep learning-based counting and body-area measurement enable sensitive population-level endpoints in the Collembola reproduction test

This paper develops an automated system using deep learning to count and measure body area in springtails, providing sensitive population-level endpoints for ecotoxicological assessments.

Predicting cell-state-specific gene regulation from DNA across species

Bolero predicts cell-state-specific gene regulation from DNA across species by integrating cell-state embeddings and biological variables.

ALLCAPS: a scalable and accurate deep-learning method for in silico pneumococcal serotyping

ALLCAPS is a deep-learning tool for in silico pneumococcal serotyping that outperforms existing methods and identifies novel serotypes.

Bypassing Current Biosecurity Screens with AI Designed Proteins and Closing the Gap with Edge-AI Functional Screening

The paper introduces Capiti, an edge-deployed system that detects pathogenic protein function during DNA synthesis by analyzing valve control signals, outperforming existing sequence-based screening tools.

ASD-FER: An Annotated Multi-Class Benchmark and Comparative Study for Facial Emotion Recognition in Children with Autism Spectrum Disorder

ASD-FER introduces a benchmark dataset for facial emotion recognition in children with ASD and evaluates multiple models, finding ViT achieves the highest accuracy on a held-out set.

What does domain-specific pretraining buy? A 170,000-parameter plant peptide language model against the ESM-2 scaling curve

The paper trains a domain-specific plant peptide language model and compares its performance to generalist models, finding domain pretraining provides significant efficiency gains.

Network organization of protein binding-site landscapes across scales

The paper uses network analysis of protein binding sites to reveal organizational patterns across scales, showing both diversification within protein categories and local similarities across boundaries.

Design of enantioselective photoenzymes by scaffolding mirror image catalytic sites

This paper designs enantioselective photoenzymes using deep learning methods to precisely position a photosensitizer and substrate, achieving high stereocontrol and regiocontrol.

MorphCell learns interpretable shape representations of 3D cellular morphology

MorphCell learns interpretable 3D cellular morphology representations from point clouds, transferring across imaging modalities and biological contexts with improved classification performance.

Physics-informed neural network for predicting myocardial mechanics from in vivo physiological measurements

This paper introduces a physics-informed neural network to predict myocardial mechanics from physiological measurements, showing improved stress prediction and disease stratification in pulmonary hypertension.

MIEN: a deep-learning model for pocket-conditioned glycan pose and conformation discrimination

MIEN improves glycan pose ranking by distinguishing experimentally observed conformations from non-native ones in protein-carbohydrate interactions.

NeuroPilot: Longitudinal intravital neural imaging across scales and modalities through an agentic platform

NeuroPilot converts heterogeneous intravital neural recordings into stable population activity maps, enabling behavioral decoding and structured neural dynamics analysis across behavioral states.

RDEmbed: A Sentence-Embedding Framework for Rare Disease Phenotype Representation and Phenotypic Similarity Analysis

RDEmbed uses sentence embeddings to analyze rare disease phenotypes from clinical text, outperforming ontology-based methods in similarity analysis and clustering.

Stain Normalisation Does Not Consistently Improve Matched-Location Retrieval with Pathology Foundation Models

This study tests if stain normalization improves retrieval with pathology foundation models, finding no consistent benefit and potential harm from Macenko normalization.

TiSAM: fusing spatial and histopathology foundation models for whole-slide image segmentation

TiSAM improves whole-slide image segmentation by fusing spatial and histopathology features, achieving high accuracy across diverse datasets and scanners.

Determining cellular phylogeny and migration at high resolution in spatial genomics

StarTrak infers cellular migration and phylogeny by integrating somatic variants, spatial data, and transcriptomics at single-pixel resolution in spatial genomics.

RiboGRAM learns multi-scale non-coding RNA grammar for structural and regulatory inference

The paper introduces RiboGRAM, a model that learns non-coding RNA sequence grammar for structural and regulatory inference across multiple organisms.

Distinct activation mechanisms underlie ligand efficacy at a GPCR

This study uses deep mutational pharmacology to show how different residues in the -opioid receptor determine the efficacy of opioids by shaping distinct activation mechanisms.

Relation-Aware Graph-Hypergraph Networks for Antibody Sequence-Structure Co-Design

This paper introduces RA-GHN, a framework for antibody sequence-structure co-design that captures multi-body interactions and structural constraints, outperforming existing methods.

qcGEM: a graph-based molecular representation with quantum chemistry awareness

qcGEM introduces a quantum-chemistry-aware graph-based molecular representation that improves performance and interpretability in molecule-related tasks.

Distance-AF Multimer: A Multi-Step Approach for Protein Complex Structure Modeling with User-Defined Distance Constraints

Distance-AF Multimer improves protein complex structure prediction by incorporating user-defined residue distance constraints, achieving competitive or better performance than existing methods.

CAPO: Compute-Aware Automated Protein-Model Optimization

CAPO automates and selects protein-model adaptations by measuring feasibility and cost before training, improving performance while reducing compute usage and failures.

DeepD enables efficient genome modeling and controllable long-context sequence generation across domains of life

DeepD is a genome foundation model that enables accurate sequence-to-function prediction and controllable genome-scale sequence generation across domains of life.

A relational architecture of the human genome

The paper systematically perturbs the human genome to reveal a relational network of genomic dependencies, uncovering 'relons' that reflect functional and evolutionary constraints.

HighFold-Lasso: Topology-Guided Prediction of Lasso Peptide Structures Using AlphaFold3

HighFold-Lasso predicts lasso peptide structures using topology-guided methods, achieving perfect topology success and low RMSD on a test set of 52 structures.

Sequential Sampling as a Computational Principle of Episodic Memory

This paper proposes that sequential gaze sampling constructs episodic memory representations through a computational model showing how sampling history shapes memory storage and retrieval.

BioMINTREL expands biological problem solving through cross-disciplinary representations

BioMINTREL enhances biological analysis by integrating cross-disciplinary methods, improving agreement with experts and uncovering new insights from existing data.

Self-extending DNA recording with stochastic and programmed writing

This paper introduces a method for self-extending DNA recording that combines stochastic and programmed writing, enabling the accumulation of cellular histories with joint cellular and tissue readout.

What comes after de novo? Automated lead optimization of proteins with CRADLE-1

CRADLE-1 automates protein lead optimization, improving speed and efficiency compared to rational design methods.

Systematic comparison of color representations between humans and deep neural networks: towards predicting human color perception in a vast color space

This study compares how different learning paradigms enable deep neural networks to predict human color perception across a vast color space.