Discover, rate, and discuss biology preprints.

The paper introduces Q-FSNet and Q-DirichNet, sparse Bayesian neural networks for identifying physiological sweet spots in metabolomic data from the Canadian Longitudinal Study on Aging.

scDeepVariant improves germline variant calling in scRNA-seq by integrating population allele frequency data, outperforming existing methods in precision, recall, and rare variant detection.

The paper shows that high accuracy in predicting enzyme kcat fold-changes from sequence data often results from data biases, not true mechanistic understanding, and introduces a bias-aware framework to improve predictions.

Shiny AMMOA provides an interactive platform for analyzing multi-omics datasets from murine aging studies, enabling integrative analysis and visualization of transcriptomic, proteomic, and metabolomic data.

The paper introduces a deep-learning framework combining LSTM, GNN, and attention for maize yield prediction by modeling genotype-environment interactions, showing improved performance with a novel architecture.

TRINUS decouples cell lineage identity from niche pressure to uncover interaction syntax in spatial transcriptomics data, showing improved clustering and domain detection across biological systems.

The paper develops an 8-gene signature to predict bevacizumab resistance and prognosis in colorectal cancer, revealing an immunosuppressive microenvironment associated with poor outcomes.

SNooPy is a statistical framework for long-read metagenomic variant calling that outperforms existing methods and is available on GitHub.

The paper introduces MaxSBM, a MaxQuant module for identifying sequence-based modifiers like SUMO, improving peptide identification rates and spectral annotation.

The paper identifies two compounds targeting NLPA, a lipoprotein involved in biofilm formation in Acinetobacter baumannii, showing antibacterial and biofilm inhibition activity in vitro and reduced intracellular survival in macrophages.

Pan1C is a pipeline that builds chromosome-level pangenome graphs, improving alignment and variant calling by integrating multiple strategies and benchmarked across different species.

ProMoNet improves drug discovery by predicting protein-ligand binding affinity and kinetics using sequence-based pre-training without 3D structural data.

This paper introduces a latent-space calibration method for rare disease ODE models using synthetic data, improving parameter recovery and trajectory prediction from sparse clinical observations.

This paper designs cell-specific small molecule cytotoxics using AI, combining screening, machine learning, and reinforcement learning, resulting in validated compounds with improved hit discovery rates.

ChatDIA uses a zero-shot large language model to automate targeted DIA proteomics analysis with high accuracy and transparent rationales.

DENcode estimates transmission probabilities between dengue cases using genetic and epidemiological data, revealing key transmission roles and network structures.

CharacTERT classifies hTERT missense variants using machine learning models that integrate sequence and structure-based features, achieving high accuracy and sensitivity in predicting pathogenicity.

TREAD improves repeat annotation by using protein embeddings instead of sequence profiles, outperforming HMMER in low-data and high-divergence scenarios.

HiCP2GAN enhances low-resolution Hi-C data using a pretrained Vision Transformer as a discriminator, improving resolution over existing methods.

This paper introduces MPRabc, an MPRA-informed model that improves enhancer-gene interaction prediction across cell types.