Discover, rate, and discuss biology preprints.

This paper introduces a method to characterize dolphin echolocation clicks using a sparse modal representation, showing accurate reconstruction and robustness under various environmental conditions.

The paper benchmarks viral prediction tools in low-abundance metagenomes and finds that combining tools improves viral detection while reducing false positives.

The paper evaluates how segmentation accuracy affects biological fidelity in spatial pathology, finding that segmentation errors significantly impact tumor-immune interactions despite similar segmentation performance.

The paper curates an accessible human surfaceome using an LLM pipeline, outperforming existing databases and identifying new candidate surface proteins.

muDM is a multi-resolution data model for spatial bioimaging that enables querying of vector geometries with associated molecular data across resolutions, improving classifier performance for neuron meshes.

This study improves classification of prodromal Parkinson's Disease using gut microbiota data and clinical features, achieving an AUC of 0.917 with integrated multimodal data.

The paper evaluates updated 16S rRNA sequencing methods and databases for improved species-level classification of the urinary microbiome.

scOmnom provides a scalable, semi-automated workflow for single-cell RNA-seq analysis, enabling auditable clustering-resolution selection and annotation refinement across large datasets.

DCSMAP improves read alignment accuracy and efficiency by selectively applying pangenome graph alignment to reads with unreliable linear placements.

The paper analyzes the geometry of breakpoint medians and anti-medians in random signed multichromosomal genomes, showing why medians favor single inputs and what balanced compromises cost.

PeakLocus assigns RNA features to transcript intervals with high accuracy, outperforming existing tools in benchmark tests across multiple species.

MIEN improves glycan pose ranking by distinguishing experimentally observed conformations from non-native ones in protein-carbohydrate interactions.

This study identifies factors governing cysteine detection in chemoproteomics using machine learning and statistical filtering, showing that peptide properties and abundance are key determinants.

The paper audits a benchmarking framework for single-cell perturbation models, revealing flaws in metric interpretation and calibration, and identifies a specific calibration failure in split-conformal prediction.

CytoRipple identifies tissue-level polarization vector fields from spatial multi-omic data to reveal multi-cellular structures and their functional associations.

This study uses supervised machine learning to predict lifespan-extending compounds in C. elegans by comparing different feature representations and models.

This paper introduces MetaKG, a metabolite-centric knowledge graph with self-supervised learning that improves representation learning and uncovers biochemical relationships.

HyARQ is a Snakemake workflow for allele-specific read counting and quality control across multiple assays in F1 hybrids, showing accurate recovery of parental proportions.

The paper introduces an interactive Data Integrator for uniform processing and integration of bulk RNA-seq datasets from the MorPhiC program and external sources.

The paper identifies robust and clinically relevant gene signatures in breast cancer by integrating two RNA-seq workflows and validating them across multiple datasets and analyses.