Discover, rate, and verify biology preprints.

Dual-graph deconvolution of enzyme fitness: an interpretable framework to identify and explain catalytic modulators

The paper introduces a hybrid framework to isolate stability-independent activity effects in enzymes using dual-graph deconvolution, validated with experimental data.

PatientPheX: A Full-Text Biomedical Corpus for Patient-Specific Phenotype Information Extraction

PatientPheX is a biomedical corpus for extracting and normalizing patient-specific phenotype information from full-text articles, achieving high entity and relation extraction performance.

An integrated data exploration engine of knowledge, expression and genetic association for human transposable elements

TE-Dee integrates TE knowledge, expression, and genetic associations to provide a structured resource for TE biology and AI model training.

A public-data reproducibility benchmark of a 2.5-D U-Net with a slice-level topology term for type-B aortic dissection segmentation

The paper benchmarks a 2.5-D U-Net with a slice-level topology term for type-B aortic dissection segmentation using public CTA data, finding no consistent benefit from the topology term.

Biological and technical spatial variation in spatial transcriptomics

The paper develops a model to distinguish biological from technical spatial variability in spatial transcriptomics data and introduces the stipple software for analysis.

2,234 attempts to harmonize a multi-platform B-cell lymphoma bulk transcriptomic dataset: ComBat is superior by cross-batch prediction quality

The paper benchmarks 2,234 transcriptomic harmonization approaches on B-cell lymphoma data, finding ComBat excels in cross-batch prediction quality.

Pre-binning correction of chimeric contigs reduces reference-discordant gene annotations in near-complete metagenome-assembled genomes

The paper introduces DeepMM, a deep learning method to correct chimeric contigs in MAGs, improving gene annotation accuracy and reducing reference-discordant genes and virulence-factor calls.

Programmable design of functional RNAs through multimodal generative learning

RNAverse, a multimodal generative model, designs functional RNAs with high performance across tasks, outperforming existing methods and enabling applications in gene therapy and vaccine development.

Longitudinal treatment selection under irreversible discontinuation: History-aware prediction and trajectory-aware learning in peanut-allergy prevention

This paper introduces TreatNet, a framework for longitudinal treatment selection under irreversible discontinuation, showing improved accuracy in peanut-allergy prevention decisions.

NoiTox: A Machine Learning Strategy for Peptide Toxicity Prediction Reveals Label Noise in Benchmark Datasets

NoiTox presents a multi-model approach for peptide toxicity prediction, revealing label noise in datasets and showing that physicochemical descriptors are informative.

PDBe-SIFTS: an open-source tool for Structure Integration with Function, Taxonomy, and Sequences, featuring improved alignment, scoring scheme, and accelerated search

PDBe-SIFTS improves residue-level sequence-structure mapping with faster search, better scoring, and accurate alignment, comparable to expert curation.

Multimodal phenotyping reveals plant-food-related dietary information in serum Raman spectra of community-dwelling Japanese adults

The paper links dietary patterns with serum Raman spectra and metabolomics in Japanese adults, showing dietary information is distributed across complementary phenotypic layers.

Protein novelty, sequence similarity, and negative sampling controls reveal benchmark specific discrimination in protein protein interactionprediction

The paper evaluates PPI predictors using controlled benchmarks and finds that performance varies with protein novelty and negative set characteristics, not generalizable to sequence-based models.

scMTG reconstructs single-cell temporal dynamics with Markov transition generators

scMTG reconstructs single-cell temporal dynamics from unpaired time-series data by learning cell representations and transitions, outperforming existing methods in interpolation and transition matrix inference.

E3 Ligase Ligandalyzer: A Structure-First, Ligand-Centric Platform for E3 Ligase Recruiter Selection

E3 Ligase Ligandalyzer provides a platform for exploring E3 ligase-recruiter complexes, enabling analysis of chemical and structural data to support degrader design.

NCI-Profiler: automated protein-ligand interaction detection and visualization

NCI-Profiler detects and visualizes nine types of protein-ligand non-covalent interactions with high accuracy and speed, outperforming existing tools in several categories.

From Closed-Set Annotation to In-Context Discovery: Biologically Grounded LLMs for Cellular Profiling

CellScribe is a generative framework that unifies cell identity and state by using a language model to generate cell type annotations and meta-program profiles, outperforming baselines and identifying new cell types through in-context learning.

Computational Prediction of Recombinant Expression Efficiency for Industrial Enzymes: An Integrated Co-Translational Sequence Analysis Platform

The paper presents a computational platform to predict recombinant expression efficiency of industrial enzymes by analyzing sequence features and co-translational processes, showing preliminary validation on solubility and other metrics.

Bayesian phylogenetics with incomplete lineage tracing data

This paper introduces a Bayesian framework to improve phylogenetic inference from sparse lineage tracing data by modeling barcode loss, showing it enhances accuracy and recovers higher growth rates than standard filtering methods.

SVC: A Vision Transformer-based Spatially and Virtually Embedded Cell Model for Deciphering Subcellular Spatial Transcriptomic Heterogeneity

SVC predicts subcellular transcript localization and co-organization using a Vision Transformer model, revealing disease patterns and spatial biology not accessible at the cell level.