Discover, rate, and discuss biology preprints.

Roughly 25× short-read depth polishes nearly all residual errors in Nanopore-only bacterial assemblies; new '--careful' modes in Pypolca and Polypolish largely avoid introducing low-depth errors.

Pretrained on 57 million dissociated and 53 million spatial human and mouse cells, a 49.3M-parameter transformer predicts niche, region, neighborhood composition and density, and carries spatial labels onto scRNA-seq.

SQANTI3 extends long-read transcriptome quality control with transcript-end subcategories, a short-read TSS ratio, rules or random-forest filters and a rescue module, raising SIRV-benchmarked F-score stepwise.

SpatialData, a Zarr/OME-NGFF-based format plus Python library, unifies images, labels, points, shapes and tables across spatial omics platforms with coordinate alignment, lazy loading and cross-modal aggregation.

A uniformly processed set of 2,440,377 bacterial and archaeal genomes with search tools and annotations, plus an AI-mined peptide set yielding ATB20, active in a mouse skin infection model.

Likelihoods from Progen2 and ESM2 protein language models are systematically higher for sequences from some species, tracking evolution-weighted database sequence counts, and pull likelihood-guided designs away from extremophile properties.

FCS-GX, a hashed k-mer cross-species aligner with a 709 Gbp curated database, screens most genomes in 0.1-10 minutes and flagged 36.8 Gbp of contamination in GenBank.

A Local Interaction Score (LIS), computed from AlphaFold-Multimer predicted aligned error, separates curated fly and human protein interactions from controls better than ipTM, recovering small or flexible interfaces.

SigProfilerAssignment refits known substitution, doublet, indel and copy-number mutational signatures to single samples and probabilistically to individual mutations, outperforming four refitting tools on 2,700 synthetic genomes.

Bounding boxes from a transformer detector let the Segment Anything Model segment cells automatically; one generalist model beats generalist Cellpose and matches inter-annotator agreement across mammalian, yeast and bacterial images.

Pairing each residue with its Foldseek 3D-structure token and pre-training an ESM-2-sized model on about 40 million AlphaFold2 structures beat ESM-2 and structure-based baselines on variant-effect and function tasks.

CollecTRI-derived signed regulons, spanning 1,183 transcription factors (TFs) and 45,856 TF-gene links, outperform DoRothEA, ChEA3, RegNetwork and Pathway Commons at identifying perturbed TFs in KnockTF data.

skani estimates average nucleotide identity (ANI) from sparse FracMinHash k-mer chains, staying robust to incomplete, fragmented metagenome-assembled genomes while running over 20 times faster than FastANI.

A Foldseek-based linear-time structural clustering groups 52 million AlphaFold DB representatives into 2.27 million non-singleton clusters; 31% lack known structural or domain annotation.

With active-state templates and shallow ortholog alignments, AlphaFold2 models all 437 human catalytic kinase domains as active; 80% of 130 benchmarked kinases are under 1 Å activation-loop RMSD.

QuantUMS, a machine-learning module in DIA-NN, weights MS1 and fragment signals by quality-derived variance estimates, improving DIA precision, removing ratio compression in high-accuracy mode, and scoring each quantity's reliability.

Combining pseudobulk differential expression, cell-type specificity, sample prevalence and NicheNet-v2 ligand activity, MultiNicheNet ranks condition-specific ligand-receptor-target signaling in multi-sample scRNA-seq with batches or multifactor designs.

Foldseek's 3Di tertiary-interaction alphabet enables structure search four to five orders of magnitude faster than Dali and TM-align at 86% and 88% of their respective sensitivities.

ENCODE's containerised Workflow Description Language pipelines for ChIP-seq, ATAC-seq, DNase-seq, RNA-seq variants, Hi-C and bisulfite sequencing run on cloud or local clusters, with documented compute costs and provenance.

Cell Painting images and profiles for 115,796 compounds in U2OS cells from 12 laboratories; 15.2% of analyzed compounds are phenotypically active, and cross-laboratory replicates cut reference-compound activity by 34%.