Discover, rate, and discuss biology preprints.

ATHILAfinder detects ATHILA LTR retrotransposons in plant genomes with high accuracy, outperforming existing tools in recovery and precision.

aaKomp provides a fast, memory-efficient, alignment-free method for genome completeness assessment, outperforming existing tools in speed and scalability.

A ProtT5-embedding, ConvNext-plus-attention ensemble predicts human pathogenic capacity from whole bacterial genomes, outperforming prior predictors on species absent from training and flagging influential proteins.

This study benchmarks deep learning methods for predicting antibody-antigen complex structures, finding that increased sampling improves accuracy but model selection remains challenging, with AlphaFold3 performing best overall.

MegaPX is a fast and space-efficient tool for classifying de novo peptide sequences against large protein databases using k-mer-based methods and Bloom filters.

PepBERT is a lightweight peptide language model that outperforms ESM-2 on most peptide-related tasks, enabling better representation of bioactive peptides for downstream applications.

The paper introduces geometry-based descriptors to predict enzyme-substrate interactions with high accuracy and interpretability, outperforming existing machine learning methods with minimal training data.

The paper uses long-read metagenomic sequencing to analyze urban soil samples, revealing novel microbial lineages and extensive functional diversity.

This study evaluates uncertainty quantification methods to improve reliability of genomic language models under distribution shift in genomic and metagenomic tasks.

This study identifies molecular signatures for frailty using transcriptomic data from two datasets, highlighting vitamin D metabolism and inflammatory pathways as key features.

This paper evaluates a genome language model for predicting bacteriophage hosts using embeddings, finding it effective for narrowing host candidates and capturing broad evolutionary relationships when combined with existing methods.

VEFill is a gradient boosting model that imputes missing DMS scores across protein domains, achieving strong predictive performance and generalization to new proteins.

HViLM is a foundation model for viral genomics that predicts pathogenicity, transmissibility, and host tropism across multiple viral families with high accuracy.

LazySlide provides an open-source Python package for whole-slide image analysis and multimodal integration, enabling efficient histopathology workflows with minimal setup.

The paper introduces UCASpatial, an ultra-precision spatial transcriptomics deconvolution algorithm, to improve the identification of immune cell subpopulations and their functional roles in tissue microenvironments.

This paper develops an EEG-based framework using CNNs and mel-spectrograms to detect schizophrenia, identifying brain lobe-specific performance and explainability with XAI techniques.

GenBio-PathFM is a foundation model for histopathology that achieves state-of-the-art performance with less training data by using a diverse data curation pipeline and a dual-stage learning strategy.

ISdetector maps insertion sequences and detects structural variations from short-read data with high accuracy and scalability.

The paper benchmarks five machine learning-based variant annotation methods for rare variant association testing across 350,377 UK Biobank participants, finding CADD provides best signal separation and GPN-MSA highest enrichment for loss-of-function intolerant genes.

ChiMER detects enhancer-exon fusion transcripts by integrating chromatin architecture into splicing graphs, improving sensitivity over conventional methods.