benchmark study

Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear baselines

The paper benchmarks deep learning models against linear baselines for predicting gene perturbation effects and finds they do not outperform simple models.

Effects of complex whole-body movements on EEG activity: a scoping review

This scoping review synthesizes studies on how complex whole-body movements affect EEG activity, finding increased theta and alpha activity in specific brain regions.

DALE-Eval: A comprehensive cell type-specific expression deconvolution benchmark for transcriptomics data

DALE-Eval benchmarks cell type-specific expression deconvolution methods with new metrics and performance insights.

BioLLM: A Standardized Framework for Integrating and Benchmarking Single-Cell Foundation Models

BioLLM integrates and benchmarks single-cell foundation models, revealing scGPTs' robust performance and highlighting limitations of scBERT in gene-level tasks.

Revealing reassortment in influenza A viruses with TreeSort

TreeSort detects influenza A reassortment in thousands of genomes via clock-based tests; avian H7 reassorts most, and HA, NA, MP rarely reassort in human and H5N1.

A comprehensive data repository of environmental toxicants exposed mouse epigenomes from the TaRGET II Consortium

The paper introduces a data repository containing epigenomic data from mice exposed to various environmental toxicants, enabling research on disease susceptibility.

Tier-based standards for FAIR sequence data and metadata sharing in microbiome research

This paper proposes tiered badges and an evaluation tool to assess data sharing compliance in microbiome research and finds that nearly half of publications fail to meet minimum sequence data availability standards.

Universal orthologs infer deep phylogenies and improve genome quality assessments

The paper improves genome assembly quality assessments by incorporating evolutionary history and introduces a curated set of BUSCOs and phyca software for phylogenetic analysis.

Can large language models reliably extract human disease genes from full-text scientific literature?

This paper benchmarks three LLMs for extracting gene-disease-phenotype relationships from biomedical texts, finding GPT-4o achieves 88.8% accuracy using zero-shot prompting.

What's next for Indian Ornithology? 101 key research questions

This paper presents 101 research questions for advancing Indian ornithology, identified through expert curation and scoring based on generality, novelty, and relevance.

The NIH BRAIN Initiative Impacts in Systems and Computational Neuroscience and Team-Scale Research, 2014-2023

This report evaluates the impact of the NIH BRAIN Initiative on systems, integrative, and computational neuroscience research from 2014 to 2023, comparing it to other funding models.

Benchmarking of tools for resolving the plasmidome from short-read assemblies for Klebsiella pneumoniae

The paper benchmarks 44 tools for plasmidome reconstruction from short-read assemblies in Klebsiella pneumoniae, finding PlaScope and MOB-suite as top performers.

Critical assessment of the ability of Boolean threshold models to describe gene regulatory network dynamics.

This study evaluates two threshold formalisms for Boolean gene regulatory networks, finding that while Ising rules match fewer functions exactly, they provide better average agreement with biological dynamics.

Cancer Alpha: A Production-Ready AI System for Multi-Modal Cancer Genomics Classification

Cancer Alpha is an AI system for multi-modal cancer genomics classification that integrates diverse datasets and production infrastructure, achieving high accuracy in synthetic benchmarks.

A comparison of tools and read depth criteria for genuine single nucleotide polymorphism identification in ancient maize samples

This study compares tools and read depth criteria for identifying genuine SNPs in ancient maize samples, finding Rm5nt and angsd with 5X depth as optimal.

Integrated Multi-omics Prioritizes Ribosomal-targeting Omacetaxine Mepesuccinate for Optic Neuritis Therapy

The paper identifies 46 optic neuritis-associated genes and prioritizes omacetaxine mepesuccinate as a repurposed therapy targeting ribosomal proteins.

A contextualised protein language model reveals the functional syntax of bacterial evolution

Trained on over 1.3 million bacterial genomes represented as ordered protein sequences, the transformer foundation model Bacformer predicts protein interactions, operons, essentiality, and phenotypes while generating trait-conditioned synthetic proteomes.

Rapid terabase-scale simulation of realistic metagenomes for experimental design and pathogen detection with RandomReadsMG

RandomReadsMG simulates realistic metagenomes for benchmarking and AI/ML training, offering faster performance and accurate depth distribution modeling.

Visible neural networks for multi-omics integration: a critical review

This paper reviews 86 studies on biologically informed neural networks for multi-omics integration, highlighting trends and challenges in terminology, tools, and evaluation methods.

Open Benchmarking for Cell-Based Multiscale Models: Lessons from a Community Initiative

This paper presents a community-driven benchmarking initiative for cell-based multiscale models, highlighting variations in tool implementations and advocating for standardized benchmarks and interoperability.

The extracellular vesicle transcriptome provides tissue-specific functional genomic annotation relevant to disease susceptibility in obesity

The paper identifies obesity-related transcriptional changes in circulating extracellular vesicles and links them to adipose tissue epigenetics and inflammatory disease risk.

A systematic assessment of phylogenomic approaches for microbial species tree reconstruction

This study evaluates four microbial phylogenomic methods on simulated and real data, finding that accounting for gene tree error improves species tree accuracy, though all methods face biases in complex datasets.

OrthoFinder: scalable phylogenetic orthology inference for comparative genomics

OrthoFinder v3 splits orthogroups at duplications predating the species-tree root, raising OrthoBench accuracy, and assigns extra species to a core run, finishing 1024 species in 128 hours.

Beyond protein functions: evaluating completeness, coherence, and consistency of genome-scale function annotations

The paper evaluates genome-scale protein function annotations for completeness, coherence, and consistency, finding that computational methods fail to produce biologically plausible annotations despite high individual protein accuracy.

Populus VariantDB v3.2 facilitates CRISPR and Functional Genomics Research

Populus VariantDB v3.2 improves CRISPR and functional genomics research by enabling precise gRNA and primer design through querying sequence polymorphisms in poplar genomes.

Scientifically Uninformed Legislation System Undermines Ecological Sustainability of the Yangtze River

The paper evaluates how scientifically uninformed legal practices in China's Yangtze River region negatively impact ecological sustainability through ineffective fish release remedies in environmental litigation.

Comparative Evaluation of Glycoproteomics Software for Rare Glycopeptide Identification

This study evaluates five glycoproteomics software tools for identifying rare sialic acid-containing glycopeptides, finding that each tool has unique strengths and limitations in detection accuracy and reliability.

Benchmarking single-cell multi-modal data integrations

This paper benchmarks 40 single-cell multi-modal integration algorithms across paired, unpaired, and mosaic datasets to evaluate usability, accuracy, and robustness.

Systematic cross-study assessment of RNA-Seq experimental workflows for plasma cell-free transcriptome profiling

The paper evaluates RNA-Seq workflows for plasma cell-free transcriptome profiling, finding that technical factors dominate variability over biological differences.

Benchmarking Imputed Low Coverage Genomes in a Human Population Genetics Context

This study benchmarks GLIMPSE imputation on low coverage genomes from underrepresented populations, showing improved performance when using target population genetic information.