Discover, rate, and discuss biology papers.

This paper introduces GeneEmbedLLM, a framework using large language models to generate gene embeddings from multi-source textual annotations, achieving competitive performance in various downstream tasks.

TideCluster is a pipeline for clustering and classifying tandem repeats across genomes, demonstrated on Arabidopsis and Solanum species.

scPolyA provides a framework for analyzing alternative polyadenylation in single-cell and spatial transcriptomes, revealing APA patterns in cancer tissues.

This paper evaluates how much improvement in cross-cancer survival prediction is due to model selection methods versus actual concordance gains, finding evaluation effects can be as significant as reported gains.

AOPTK automates toxicological literature analysis using AI, achieving high accuracy in identifying chemical-biological effect relationships.

mFrag improves lung cancer early detection using fragmentomics by integrating DNA methylation data and combining two models to enhance prediction accuracy.

BEHAV3D-Explorer provides a user-friendly platform for analyzing 3D live-cell imaging data using natural language guidance to select analysis strategies and configure workflows.

BioMeme predicts cell responses to new perturbations by recalling measured experiments, outperforming neural models in accuracy and speed.

Ok-k-mer selects optimal k-mer length ensembles to improve genomic distance estimation by maximizing correlation with phylogenetic branch distances.

GraffiTE v1.1 is a Nextflow pipeline that genotypes structural variants, annotates polymorphic mobile elements, and maps epigenetic data to transposable elements using pangenome graphs, revealing hypomethylated elements and age-related demethylation.

The paper introduces a PyTorch-based tool for fast proteome-scale local frustration analysis, achieving 70-13,000x speedup over the original FrustratometeR.

This paper introduces efficient methods to update colored k-mer indexes without rebuilding them from scratch, enabling dynamic maintenance of large genomic collections.

The paper analyzes six-residue Omega-loops in protein complexes, showing their structural constraints, sequence conservation patterns, and functional significance through sequence design and mutation mapping.

MELD predicts SNV functional effects by directly training on wild-type versus mutant differences, outperforming post-hoc in-silico mutagenesis methods with fewer parameters.

RabbitCD-HIT clusters billion-scale protein sequences efficiently, achieving faster performance than CD-HIT on large datasets.

StableProt predicts protein thermostability and organismal optimal growth temperature using a structure-aware deep learning framework, achieving high accuracy and robustness across temperature regimes.

This paper introduces SSAF-T, a segmentation-free framework for cell-type inference in spatial transcriptomics that avoids cell segmentation and enables accurate cell-type-guided segmentation.

This paper introduces scJDO2, a method to separate score and flow operators from single-cell time courses, evaluating their ability to identify unstable cells and bifurcations in hematopoiesis.

DisCell identifies and separates spillover from niche responses in spatial transcriptomics data, confirming a previously reported finding about esophageal cancer cells.

This paper introduces IsoVita, a framework for learning interpretable compositional features from continuous covariates, showing improved detection of temporal isoform variation in human brain RNA-seq data.