Discover, rate, and discuss biology preprints.

This study benchmarks 62 text summarization methods on biomedical abstracts, finding that general-purpose large language models outperform domain-specific and reasoning-oriented models.

PeptiVerse provides a unified platform for predicting therapeutic peptide properties using both amino acid sequences and SMILES-based representations, supporting early-stage development and generative design.

CurateMake is a workflow for harmonizing and validating ITS reference databases, improving taxonomic coherence and error correction through nomenclatural standardization and phylogenetic validation.

The paper evaluates frozen DNA encoders' ability to recover protein-family labels and GenePT embeddings from coding sequences and TSS contexts, finding limited success in gene-function prediction.

This paper benchmarks Evo 2's ability to encode viral genome architecture, finding it captures higher-order features beyond sequence composition.

This study shows that synthetic cannabidiol reduces heart failure progression by modulating immune and endothelial-to-mesenchymal transition pathways in mice.

SHINE integrates spatial transcriptomics and metabolomics data to identify microenvironments, outperforming existing methods in segmentation and biomarker co-localization.

This paper re-analyzes a foundation model's checkpoints to show biological continued pretraining improves general knowledge, code generation, and reasoning without catastrophic forgetting, but subsequent fine-tuning reverses these gains.

Phylogenize2 links microbial gene families to abundance phenotypes across diverse environments, identifying genes associated with specific ecological roles.

CAPHEINE automates evolutionary analysis of pathogen sequences, performing selection dynamics and positive selection studies.

The paper designs a multi-epitope vaccine against M. tuberculosis using computational methods, showing strong structural quality and simulated immune responses.

The paper identifies IKZF1 as a key microglial transcription factor in Alzheimer's disease using single-cell data and virtual screening, with potential drug repurposing candidates.

The paper introduces a method to prioritize targets in Perturb-seq by assessing the dependability of perturbation effects using generalizability theory, improving target identification reliability.

OmiXAI is an ensemble XAI pipeline for deep learning in omics data that reduces feature sets and benchmarks XAI methods on Z-DNA prediction.

This paper introduces a 4D tensor decomposition method to integrate transcriptome, Ribo-seq, and proteome data, identifying gene clusters related to ribosome stacking and translational buffering.

The paper develops a new statistical method for metagenomic gene-level analysis that reduces false positives and identifies a novel transcription factor linked to liver cirrhosis.

Lemonite integrates transcriptomics and metabolomics data to identify regulatory metabolites acting on gene modules in glioblastoma and inflammatory bowel disease.

This study uses reinforcement learning to optimize antibiotic prescribing policies under antimicrobial resistance dynamics, showing that hierarchical RL outperforms fixed rules in complex, realistic scenarios.

The paper introduces RomicsProcessor, a reproducible-by-design tool for multi-omics and spatial-omics analysis that enforces FAIR principles and ensures computational reproducibility.