Discover, rate, and discuss biology preprints.

The paper establishes a ceiling for biological score prediction based on assay concordance and finds current predictors achieve 63% and 18% of these ceilings.

UTR-Diffusion is a diffusion model for designing 5-prime UTR and UTR-CDS junctions that optimizes translation efficiency, RNA stability, and codon adaptation while preserving sequence constraints.

The paper reconstructs a genome-wide protein-interaction network for Varroa destructor and evaluates orthology-based host gene absence calls, finding over-calling of host absence in three tests.

EcoXAI introduces a multi-agent system for biomedical discovery that integrates knowledge graphs and agentic AI to identify novel drug candidates for Alzheimer's Disease.

This study applies YOLOv8 for automated detection of livestock gastrointestinal parasite eggs and cysts in microscopic images, achieving moderate performance metrics but highlighting data limitations.

The paper develops an enrichment-free glycoproteomics method using mass defect-driven glycopeptide classification to identify sex differences in murine plasma fucosylation.

Prot-LAMBDA improves protein structure prediction by explicitly incorporating spatial relationships, achieving better performance than ESM2-3B in contact prediction and 3D structure prediction.

BindScreen improves virtual screening by using a protein-centric contrastive learning approach with asymmetric loss and batch design, achieving better performance with fewer computational resources.

This study benchmarks graph neural networks for cancer subtyping using methylation and gene expression data, finding that ChebNet performs best and simple readout layers improve performance.

This paper uses deep learning to predict relationships between 5' and 3' UTRs, finding that highly related pairs are enriched in genes involved in neural development and have distinct sequence and structural features.

UELer enables interactive exploration of multiplexed imaging datasets within Jupyter notebooks by linking image views with analysis results without requiring additional infrastructure.

eSkip2 identifies effective exon-skipping ASO target regions by integrating genomic context and splicing perturbations, outperforming exon-restricted models in prioritizing active regions.

ClustoCell identifies cell states and markers from single-cell transcriptomes by analyzing transcriptional architecture, improving accuracy and stability over existing methods.

AFP-R provides a database and predictor for antifreeze proteins, offering curated data and a machine-learning model for AFP identification.

TxNova identifies recurrent unannotated intergenic splice junctions from bulk RNA-seq alignments, providing candidate leads for experimental validation.

The paper describes a spatiotemporal mechanistic model for forecasting HPAI H5N1 and evaluates its performance in simulating outbreak dynamics and response readiness in Australia.

DeepPathway predicts pathway expression from H&E images using contrastive learning, validated on prostate cancer and brain tumor datasets.

ProtEnrich enhances protein sequence embeddings with biophysical properties without disrupting their semantic structure, improving performance on structure-related tasks.

MolJam assesses molecular dataset quality across five dimensions and shows that improving quality metrics does not always enhance machine learning performance due to trade-offs with data quantity.

The paper identifies a five-gene expression signature associated with inflammation that stratifies survival and immune states in lung adenocarcinoma across multiple cohorts.