benchmark study

GLOF: A large-scale expert-curated benchmark dataset of gain-of-function and loss-of-function missense variants

The paper introduces GLOF, a large-scale expert-curated dataset of missense variants classified as loss-of-function, gain-of-function, or neutral, to benchmark computational methods for predicting variant functional mechanisms.

SNAC-DB: An ML-Ready Database for Antibody and NANOBODY(R) VHH-Antigen Complexes with Expanded Structural Diversity and Real-World Benchmarking

The paper introduces SNAC-DB, a database with expanded structural diversity for antibody and NANOBODY VHH-antigen complexes, and benchmarks seven models showing limited accuracy and ranking issues.

An Agentic Platform for Drug Repurposing Unified across Molecular, Phenotypic, and Clinical Scales

LinkD is a drug repurposing platform that integrates molecular, phenotypic, and clinical data to prioritize drugs, showing reduced prostate cancer incidence with beta-blockers.

Re-evaluating the benefit of training scale in cell foundation models

This study evaluates how training scale affects cell foundation models by comparing performance with frozen representations and finds that larger pretraining data improves model performance.

Phylodynamic Analysis of the Global Dispersal and Spatiotemporal Dynamics of HIV-1 Subtype C

This study benchmarks phylodynamic methods to analyze HIV-1 subtype C dispersal, finding robust spatial transmission patterns despite methodological variability in temporal estimates.

A Reproducible and Extensible Benchmark of Supervised Cell Type Annotation Tools for Cytometry Data

This paper benchmarks supervised cell type annotation tools for cytometry data using a reproducible framework, showing performance varies across datasets and populations, with rare populations being a challenge.

R-loop Prediction Reveals Generalization Limits of DNA Foundation Models Beyond Regulatory Genomics

The paper evaluates DNA foundation models' ability to predict R-loop-forming sequences, finding they perform well in some settings but struggle with cross-platform and cross-species generalization.

Predicting P-glycoprotein Substrate Status Using a Pretrained Graph Neural Network: A TDC Benchmark Study

This study evaluates a pretrained graph neural network for predicting P-glycoprotein substrate status using a benchmark dataset, achieving high AUROC scores.

Plant growth forms shaping biodiversity patterns can guide conservation planning on the Qinghai-Tibet Plateau

The paper shows that plant growth forms influence biodiversity patterns on the Qinghai-Tibet Plateau, and prioritizing trees and climbers can effectively capture most diversity with minimal area.

Benchmarking Genomic Variant Calling Tools in Inbred Mouse Strains: Recommendations and Considerations

The paper benchmarks variant calling tools in inbred mouse strains, finding tradeoffs between recall and precision and recommending ensemble approaches for optimal results.

sstar2: A Python Package for S*-based Archaic Introgression Detection with Machine Learning

sstar2 improves S*-based introgression detection by using quantile regression for better threshold calibration and compatibility with modern simulations.

Leveraging the Genetics of Psychiatric Disorders to Prioritize Potential Drug Targets and Compounds

This paper uses genetic data from psychiatric disorders to prioritize drug targets and compounds, identifying genetic support for existing medications and opportunities for drug repurposing.

Positive-control Mendelian randomization highlights power constraints in disease-mortality GWAS

The paper uses Mendelian randomization to show that current disease-mortality GWAS may lack sufficient power to detect shared genetic effects, despite strong associations in case-control studies.

Measuring the Meaning of Genomic Results: Harmonization of the Metric for Case-Level Results in the CSER2 Consortium

The paper develops a standardized case-level categorization metric for genomic results across the CSER2 consortium to assess diagnostic yield and identify resolvable factors in inconclusive cases.

Benchmarking within-sample minority variant detection with short-read sequencing in M. tuberculosis

This study benchmarks seven variant callers for detecting low-frequency variants in M. tuberculosis using simulated short-read data, finding FreeBayes performs best and introducing an error model to filter false positives.

PromptBio-Bench: Benchmarking LLM-based Bioinformatics Agents for End-to-End Data Analysis

PromptBio-Bench evaluates LLM-based bioinformatics agents across 244 tasks, showing comparable performance between Biomni and ToolsGenie but declining accuracy with task difficulty.

Prevotella stercorea links gut microbiome ecology to respiratory infection protection through a host-context-dependent, species-autonomous pathway

This study identifies Prevotella stercorea's role in reducing respiratory infections through a host-context-dependent pathway, independent of microbiome richness.

Challenges and Progress in RNA Velocity: Comparative Analysis Across Multiple Biological Contexts

This study compares five RNA velocity methods across three datasets to evaluate their performance in predicting cellular states.

Comparison of AI protein structure ensemble prediction tools

This study compares AI tools for predicting protein structural ensembles, finding they generate diverse but distinct conformational distributions, with limited coverage of experimental structures and minimal response to known mutations.

Lysergic acid diethylamide reverses aging- and neurodegeneration-associated brain transcriptional programs

This study shows that LSD induces gene expression patterns opposing aging and neurodegeneration in the brain, suggesting potential therapeutic effects through transcriptional reversal.

A large-scale comparative metagenomic analysis of short-read sequencing platforms indicates high taxonomic concordance and functional analysis challenges

This study compares taxonomic and functional profiles of human gut microbiome samples sequenced on two platforms, finding high taxonomic concordance but functional discrepancies.

A Root Foundation Model for Zero-Shot Segmentation

The paper introduces a domain-specific foundation model for root segmentation, showing strong zero-shot and few-shot performance across multiple datasets.

DfE-DB: A systematic database of 3.8 million human decisions across experience-based tasks

DfE-DB compiles 3.8 million human decisions across 168 studies to enable systematic comparison of decision-making patterns and design features.

TIMS-Bench: Towards community standards for benchmarking untargeted trapped ion mobility metabolomics tools and datasets

TIMS-Bench evaluates three TIMS feature-finding tools using public and benchmark datasets, identifying strengths and weaknesses in feature detection and annotation.

Benchmarking the quantitative performance of metabarcoding and shotgun sequencing using mock communities of marine nematodes

This study benchmarks metabarcoding and shotgun sequencing for quantifying marine nematode communities using mock communities, finding shotgun sequencing more reliable for cross-taxon abundance comparisons.

A joint metabolic-integrative regime of consciousness: evidence from FDG-PET and TMS-EEG

The paper identifies a joint metabolic-integrative regime for consciousness using FDG-PET and TMS-EEG data, showing that both metabolic support and perturbational complexity are required for conscious states.

Optimizing Strain Selection for Association Studies Under Hard Cost Constraints

The paper evaluates strain selection methods under cost constraints for association studies, finding that cost-based approaches often outperform diversity-focused ones.

The Brain Encyclopedia Atlas Project (BEAP): A Literature-Synthesis-Derived Functional Atlas of the Human Brain

BEAP synthesizes functional neuroanatomy from literature to create a spatially integrated atlas of human brain regions.

dbGIST: An LLM-Assisted Multi-Omics Resource for Target Exploration and Cross-Dataset Validation in Gastrointestinal Stromal Tumors

dbGIST is a multi-omics resource for GIST research that integrates diverse datasets and provides analytical tools for target exploration and cross-dataset validation.

Benchmark Bias and Conformational Dynamics in Allosteric Site Prediction

This paper identifies a bias in allosteric site prediction benchmarks and introduces AlloDyn, a dynamic-aware framework that improves prediction performance by incorporating conformational dynamics.