benchmark study

On the robustness of scRNA-seq foundation models for plant perturbation response prediction under cross-experiment shift

This study evaluates scRNA-seq foundation models for plant perturbation response prediction, finding that performance varies significantly with evaluation protocol and that preserving gene identity improves generalization.

Genetic architecture of Alzheimer's disease is enriched in peripheral immune and metabolic programs

This study identifies peripheral immune and metabolic tissues as key contributors to Alzheimer's disease genetic risk, independent of brain-specific factors.

IUPAC Consensus References Improve Short-Read Variant Detection in Clinically Challenging Regions: A Stratified Benchmarking Study with BurdenBench

This study evaluates IUPAC consensus references for improving variant detection in challenging genomic regions using a stratified benchmarking framework called BurdenBench.

PG-LLM: Benchmarking General-Purpose Language Models for Protein Variant Ranking

PG-LLM benchmarks language models for protein variant ranking, showing they outperform many sequence-based predictors but fall short of specialized models.

From Plants to Patients: Mitochondrial Stress Signaling as a Systems Framework for Human Disease Vulnerability

This study compares plant and human mitochondrial stress signaling systems to identify cross-kingdom principles that may inform human disease vulnerability.

Task-dependent Performance of Single-cell Foundation Models under Low Supervision

This study evaluates seven single-cell foundation models across five tasks, finding varying performance depending on the downstream task and highlighting task-specific strengths.

Cryptic binding sites are detected but not ranked: coverage, conversion, and the limits of detector consensus

The paper evaluates cryptic binding site detection methods by separating coverage from ranking, showing that ranking improvements can significantly enhance recovery.

Biomarker Fidelity Score - A Quantitative Framework for Individual-Level Validation of Explainability Methods in 3D Alzheimer's Disease MRI Classification

The paper introduces a quantitative framework to validate explainability methods in Alzheimer's disease MRI classification at the individual level using biomarker-aligned spatial overlap scores.

Pharmacobehavioral space of MoSeq syllables significantly overlaps with scalar locomotion features

The paper reanalyzes MoSeq data and finds that the performance advantage of MoSeq syllables over scalar features depends on analytical methods and dataset structure, not inherent superiority.

A multi-stakeholder Delphi consensus study proposes actionable solutions to address antibody validation failures

This paper identifies barriers to antibody validation and proposes a consensus-based action plan to improve validation practices across stakeholders.

Systematic Benchmarking of AI-Based Molecular Generation Models for Structure-Based Drug Design

This study benchmarks 12 AI-based molecular generation models for drug design across 176 protein-ligand systems, finding no single method optimizes all criteria and proposing a hybrid approach.

SynthMLM: A framework for interpretable analysis and synthetic localisation data generation for SMLM

SynthMLM generates synthetic localisation data for SMLM using interpretable structural descriptors from experimental data, enabling benchmarking and testing of analysis methods.

Assessing Codon Language Models for Context-Aware Codon Optimization in Nucleic Acid-Based Medicines

This study benchmarks three codon language models for optimizing nucleic acid-based medicines, finding that while no single model excels across all tasks, they offer advantages in capturing translational context beyond codon frequency.

Benchmarking computational decontamination of ambient RNA

This study benchmarks ambient RNA decontamination methods using simulated and experimental datasets, finding that no single method is optimal across all scenarios, with CellBender, DecontX, and SoupX performing well.

Comparative evaluation of genotyping and low-pass sequencing for pharmacogenetic variant and phenotype inference

This study compares genotyping arrays and low-pass whole-genome sequencing for pharmacogenetic variant and phenotype inference, finding high concordance but improved variant coverage with LP-WGS.

FlyPredictome: A structural atlas of predicted protein-protein interactions in Drosophila

FlyPredictome scores 1.5 million AlphaFold-Multimer predictions of Drosophila protein pairs with a contact-aware integrated Local Interaction Score (iLIS), yielding residue-level interfaces enriched for phenotype-associated missense alleles.

Benchmarking Bayesian colocalization methods in validating Mendelian randomization-identified targets

This study benchmarks four Bayesian colocalization methods to validate MR-identified protein-trait associations for cardiometabolic traits, finding SharePro most frequently supports significant associations.

Functionally Oriented Genetic Analyses Reveal Potential Transcriptomic and Neurological Mechanisms of Stuttering

This study identifies transcriptomic and neurological mechanisms of stuttering using GWAS, TWAS, MR, and neuroimaging analyses of large datasets.

PlantAI: A Multi-Agent System for Plant Functional Genomics Analysis and Biological Knowledge Interpretation

PlantAI integrates bioinformatics analysis and knowledge interpretation for plant functional genomics, achieving high recall and accuracy in literature-based evidence retrieval.

Benchmarking Spectral Library Prediction Platforms for Neuropeptidomics Applications

This study benchmarks four spectral prediction platforms for neuropeptidomics, finding no single model captures neuropeptide fragmentation well and highlighting the importance of neutral loss fragments and hybrid libraries.

Human-supervised Agentic AI for Hypothesis Generation and Experimental Assistance in Drug Repurposing

A human-supervised multi-agent LLM system covered drug repurposing from knowledge graph building to assay analysis in three case studies; retrospective COVID-19 screen AUC-ROC 0.986 without predefined thresholds.

Local ancestry inference identifies robust evidence of selection in Neolithic Europe

Local ancestry methods on 176 imputed Neolithic genomes disagree on tract lengths; replicated meta-analysis supports selection at SLC24A5 and FADS1/2, while HLA signals look artefactual.

Sparse sampling and rare-variant depletion distort PCA visualizations of population structure: recovery with objective-guided manifold learning

Sparse or clustered sampling plus rare-variant filtering collapse PCA plots into triangles and three-ray patterns; objective-guided manifold learning on high-dimensional classic PC scores recovers obscured structure.

TRIDENT: a framework for robust multi-trait GWAS identifies 66 novel multi-trait osteoarthritis signals

TRIDENT is a framework for robust multi-trait GWAS that identifies 66 novel osteoarthritis signals by combining earlier and larger GWAS meta-analyses and applying translational annotation.

Benchmarking cell-type deconvolution in cross-platform transcriptomic data

This study benchmarks cell-type deconvolution methods across cross-platform transcriptomic data, finding that log-normal regression methods like SpatialDecon perform most reliably.

A dysregulated stromal remodelling programme characterises prior anti-TNF failure in ulcerative colitis

The paper identifies a stromal remodelling programme in ulcerative colitis patients who failed prior anti-TNF therapy, linking it to MAPK/EGFR signalling and suggesting MEK/EGFR inhibitors as potential treatments.

A comprehensive benchmark of transcriptome-wide fusion detection using long-read RNA sequencing

This study benchmarks fusion detection methods using long-read RNA sequencing, finding that ONT PCR-cDNA with CTAT-LR-Fusion offers the best balance of precision and recall.

Diagnostic Accuracy of Large Language Models for Rare Diseases: A Systematic Review and Meta-Analysis

This paper evaluates the diagnostic accuracy of large language models for rare diseases through a systematic review and meta-analysis, finding substantial variation in performance across benchmarks and high risk of bias in all studies.

Reliable single-cell perturbations explain and improve model performance

The paper evaluates perturbation reliability in single-cell data and shows that using reliable perturbations improves model performance and reduces training data needs.

Designing to Implement Genomics Informed ASCVD Risk Assessment: Patient and Clinician Perspectives about Identifying and Managing the Underlying Causes of Severe Hypercholesterolemia

The paper explores patient and clinician perspectives on identifying and managing the underlying causes of severe hypercholesterolemia through genomic insights.