Discover, rate, and discuss biology preprints.

Hi-cGAN predicts Hi-C interaction matrices using conditional generative adversarial networks, achieving better performance than existing methods at certain resolutions.

Atlantis integrates structural and functional residue-level data for the human proteome, enabling complex queries and analysis of protein interactions and mutations.

The paper evaluates machine learning predictors of enzyme turnover numbers (kcat) by testing their impact on downstream metabolic models, finding that accuracy varies significantly and that validation must consider system-level effects.

PocketScope retrieves binding sites across the AlphaFold human proteome using protein language model representations and achieves top performance on a benchmark.

BioIMA is a desktop tool that automates phenotypic trait extraction from biological images, achieving high accuracy and speed compared to manual methods.

The paper annotates neuropeptides and GPCRs in Gryllus bimaculatus and predicts ligand-receptor pairs using AlphaFold3 and Boltz-2, providing a resource for studying insect endocrinology.

scRep is a foundation model for single-cell representation learning that uses self-distillation in latent space, achieving strong performance without task-specific fine-tuning.

The paper develops and validates a prognostic model using morphologic intratumoral heterogeneity from histopathology slides to predict survival in primary central nervous system lymphoma.

The paper introduces Bravais Lattice Sampling (BLS), a method for detecting connected high-density regions in 3D space, achieving high accuracy with reduced computational cost compared to traditional methods.

BC-SELECT predicts treatment response in non-metastatic breast cancer using genetic interactions and gene expression data, showing significant accuracy for PARP inhibitors and immunotherapy.

The paper introduces IFACE, a method to compute geometric-chemical distances between protein surfaces and provides surface maps, showing improved performance over existing methods in classification and clustering.

GLORB improves differential expression analysis by accounting for global expression shifts, outperforming existing methods in scenarios with widespread upregulation or downregulation.

AnnFlux models single-cell perturbation dynamics using object-conditioned neural SDEs, predicting responses to unseen perturbations and combinations.

Ageas infers cell fate from static single-cell and spatial multi-omics data without time-resolved measurements, revealing spatial fate gradients in human embryos.

This study identifies high-risk nsSNPs in TSC1 and TSC2 genes using computational tools, showing they reduce protein stability and disrupt functional sites.

The paper identifies trans-eQTL hotspots in T. urticae linked to host plant fitness using a multiparental RNA-seq approach.

DAG-HEART improves breast cancer outcome prediction in underrepresented populations by integrating multi-omics data with directed acyclic graph constraints and transfer learning.

TRACTINATOR infers continental and archaic ancestry in admixed genomes using deep learning, revealing archaic introgression patterns in Latin American populations.

This study evaluates the impact of pre-trained protein language models on drug-target affinity prediction, showing that simple architectural changes can match large PLMs' performance.

This paper introduces a framework combining information geometry and UMAP to model gene expression dynamics and identify stem cell-related genes in single-cell RNA sequencing data.