Discover, rate, and discuss biology preprints.

The paper introduces a Vector-Symbolic Architecture for encoding viral pangenomes and evaluates its effectiveness in species classification, finding that a flat species-level model outperforms genus-level and hierarchical approaches.

The paper identifies Lactobacillus crispatus as a potential probiotic for UTI prevention by analyzing urinary microbiome differences between UTI patients and healthy individuals.

KaryoScope annotates genome assemblies at base-pair resolution across multiple feature classes quickly and without alignment, revealing structural variation in centromeres and subtelomeres.

MetaNet is an R package for scalable multi-omics network analysis, offering faster computation and better memory efficiency than existing tools, with applications in microbial and exposome-transcriptome studies.

cfMIND improves non-invasive disease detection by analyzing read-level methylation signals from cell-free DNA, achieving high accuracy even at low sequencing depth and across diverse cancers.

This paper benchmarks synthetic data generation methods for clinical transcriptomic cancer data, finding no single method outperforms all metrics, with Gaussian Copula showing the most balanced performance.

EMReady2 improves cryo-EM and cryo-ET maps using a Mamba-based deep learning model, enhancing quality and interpretability while reducing computational cost.

The paper investigates genomic language model representations by analyzing sparse auto-encoder features, finding they encode granular sequence syntax and conservation patterns rather than complex regulatory logic.

The paper introduces a hybrid annotation framework that improves cross-species protein similarity assignment and gene annotation in evolutionarily divergent genomes, expanding structural catalogs and improving single-cell immune cell state annotation in sea lamprey.

The paper introduces kaleidoCell, a framework for identifying transcriptional meta-programs in single-cell datasets, and applies it to glioblastoma data to uncover treatment-related signatures.

OpenBase provides a framework for high-accuracy detection of diverse non-canonical DNA bases using nanopore sequencing through standardized training data and deep learning models.

This paper introduces a Bayesian framework for inferring transmission networks from polyclonal infections, demonstrating its application to Plasmodium falciparum with high precision and recall in simulations.

Coxmos provides an interpretable survival analysis tool for high-dimensional and multi-omic data, outperforming existing methods while enhancing model interpretability.

This study evaluates the reliability of physics-informed neural networks (PINNs) for parameter recovery in the repressilator model, showing accurate trajectory reconstruction but challenges in parameter estimation due to noise and model assumptions.

TwinSAR detects chemical twins using stoichiometric analysis, improving virtual screening by identifying structurally diverse but compositionally similar compounds.

This paper develops a machine learning framework to provide personalized dietary recommendations for achieving desired gut microbial compositions using population-level cohort data.

This study shows TDP-43 binds and stabilizes DNA G-quadruplexes, enhancing chromatin looping and gene expression in K562 and HepG2 cells.

This study benchmarks long-context genome language models on biosynthetic gene clusters, showing layer selection impacts performance and revealing how models encode biological information.

This paper identifies challenges faced by novice genomic researchers with bioinformatics tools and proposes an evaluation rubric to improve software usability.

This paper introduces VIP2B, a method to extract viral features from bulk metagenomes, showing that viral data improves disease diagnosis and reveals distinct gut virome community states.