Discover, rate, and discuss biology preprints.

This paper introduces a probabilistic framework for analyzing motor states in Parkinson's disease, showing they are dynamic and associated with imaging features.

The paper introduces uncertainty-aware gene ranking methods to improve stability in coexpression network analysis, showing better recovery of reference and tissue-specific genes compared to traditional rankings.

DOMINO identifies spatial domains in spatial transcriptomic data with improved accuracy and scalability, outperforming existing methods on large datasets.

MetaHarmonizer improves metadata harmonization and addresses LLM contamination in biomedical data, outperforming existing methods on benchmark tasks.

PhenoBIC is a deep learning model that automates cell phenotyping in multiplex imaging data, outperforming manual methods and other ML approaches with an F1-score of ~0.88.

FlowBench evaluates agentic bioinformatics systems by decomposing performance into planning, fault recovery, interpretation, and output fidelity, revealing limitations in toolchain inference and fault handling.

Orion, an AI agent, automates biomedical image analysis by integrating large language models with software execution and reasoning, generating research reports and hypotheses from imaging data.

cuBayes accelerates whole-genome SNV calling using GPU acceleration, reducing analysis time from hours to minutes while maintaining high concordance with CPU results.

RICE improves causal gene network estimation in single-cell CRISPR screens by addressing confounding, count data, and high-MOI designs, outperforming existing methods in synthetic and real data.

OmicOS provides an agent-operable omics ecosystem that enables AI systems to execute and verify analyses using community methods, achieving high performance in benchmark tasks.

This paper introduces a physics-informed machine learning framework for infectious disease forecasting, showing improved accuracy in predicting case and hospitalization counts using a hybrid model of differential equations and machine learning.

This paper introduces SpaWeaver, a framework for robust spatial multi-omics integration using weak anchors, showing improved performance in handling real-world challenges and enabling new biological insights.

RetroMol converts natural product structures and biosynthetic gene clusters into shared primary sequences to enable biosynthetic similarity comparisons.

This paper introduces BetaInfer, a generative model for phylogenetic tree inference that maps unaligned sequences to trees, achieving competitive performance compared to traditional methods.

AutoZyme optimizes bioinformatics software by autonomously identifying and applying code changes to improve runtime without significant memory use.

DynamicDemiLog (DDL) is a unified sketch that estimates cardinality, similarity, frequency, and composition efficiently, outperforming existing methods in speed and accuracy.

SCALLOPS improves processing of optical pooled screens by enabling efficient alignment and analysis of large-scale image data, demonstrating superior performance over existing methods.

DMcloud improves fitting of molecular models into cryo-EM maps by local structure fitting, achieving higher accuracy and coverage than existing methods.

scIsoAgent improves long-read single-cell RNA-seq analysis by creating reproducible, isoform-resolved workflows and linking sequence data to functional hypotheses.

This paper introduces DeepUI, a physics-driven framework that reconstructs isotropic 3D fluorescence images from undersampled axial data, achieving high performance even under challenging conditions.