Experiments

Searchable full-text extractions: founding hypothesis, core claims, experimental setups, key results and statistics — pulled out of each paper as structure. Search a cell line, an assay or an entity (e.g. HUH7) and find every paper that worked with it. This corpus stands on its own: most entries carry no reproduction assessment (yet).

• Prioritization of candidate cancer genes--an aid to oncogenomic studies.

  PMID 18710882 · PMC2566894 · Nucleic acids research · 2008 · 8 claims · 8 setups

  Full-text index only

  Computational classifiers using combinations of protein conservation, gene structure, protein domains, protein interactions, and regulatory data can distinguish known cancer genes (CD/CR) from unlabelled human genes

  genes cancer interpro interactome averaged cross-validation domain

• Boosting accuracy of automated classification of fluorescence microscope images for location proteomics.

  PMID 15207009 · PMC449699 · BMC bioinformatics · 2004 · 8 claims · 8 setups

  Full-text index only

  New classifiers (SVMs, ensembles) and new wavelet-derived (Gabor, Daubechies) features improve recognition of protein subcellular location patterns over the previous neural network approach

  hela automated stepwise training-set-size

• Evolutionary trace annotation of protein function in the structural proteome.

  PMID 20036248 · PMC2831211 · Journal of molecular biology · 2010 · 8 claims · 7 setups

  Full-text index only

  ET-ranked residue clusters can be used to build 3D templates that predict GO function in enzymes and non-enzymes alike, without prior knowledge of functional mechanism.

  structural pdb pairs annotated evolutionary template support

• Predicting the pathogenicity of missense variants using features derived from AlphaFold2.

  PMID 37084271 · PMC10203375 · Bioinformatics (Oxford, England) · 2023 · 6 claims · 8 setups

  Has reproduction · 85

  AlphaFold2-derived structural features (solvent accessibility, amino acid network features, physicochemical environment, pLDDT) can be used to train a random forest classifier (AlphScore) that distinguishes proxy-benign from proxy-pathogenic missense variants.

  gnomad alphafold2-predicted alphscore proteins machine secondary physicochemical

• A modified T-test feature selection method and its application on the HapMap genotype data.

  PMID 18267305 · PMC5054219 · Genomics, proteomics & bioinformatics · 2007 · 7 claims · 4 setups

  Full-text index only

  A modified t-test ranking measure, extended to handle nominal SNP genotype data via vector transformation, can effectively rank SNPs by their discriminative capability for population classification.

  hapmap modified f-statistics informativeness

• Predicting Bone Metastasis Using Gene Expression-Based Machine Learning Models.

  PMID 34858485 · PMC8631472 · Frontiers in genetics · 2021 · 8 claims · 5 setups

  Has reproduction · 62

  A DNN model built on 34 top-ranked hub genes achieves the highest prediction accuracy (AUC 92.11%) for distinguishing primary from bone-metastasized tumors

  breast degs geo tcga microarray protein-protein machine

• Interaction profile-based protein classification of death domain.

  PMID 15189571 · PMC459208 · BMC bioinformatics · 2004 · 7 claims · 6 setups

  Full-text index only

  An SVM-based classifier using Residue Pair Interaction Profiles (RPIPs) can classify death domain superfamily members into subfamilies with 89% average cross-validation accuracy

  homology-modeled death homology putative rigid-body homology model

• A comprehensive sensitivity analysis of microarray breast cancer classification under feature variability.

  PMID 19941644 · PMC2789744 · BMC bioinformatics · 2009 · 7 claims · 4 setups

  Full-text index only

  Feature variability strongly influences breast cancer signature composition even when array platform and patient stratification are identical.

  breast preprocessed microarray two-color perturbation

• SePaCS--a web-based application for classification of seroreactivity profiles.

  PMID 17478503 · PMC1933220 · Nucleic acids research · 2007 · 8 claims · 4 setups

  Full-text index only

  SePaCS is a freely available web-based tool that trains and applies multiple classification methods (4 Naive Bayes variants, SVM with RBF kernel, LDA, DLDA) to seroreactivity profiles and outputs results as a summary table plus a detailed PDF report

  blood computational seroreactivity supervised mutual

• Gene-expression patterns in peripheral blood classify familial breast cancer susceptibility.

  PMID 26538066 · PMC4634735 · BMC medical genomics · 2015 · 8 claims · 7 setups

  Has reproduction · 83

  A multigene expression biomarker from PBMCs accurately classifies familial breast cancer (FBC) status

  pbmcs utah blood normal microarray svm-rfe svm

• Graph Random Forest: A Graph Embedded Algorithm for Identifying Highly Connected Important Features.

  PMID 37509188 · PMC10377046 · Biomolecules · 2023 · 8 claims · 6 setups

  Has reproduction · 89

  GRF identifies effective features that form highly connected sub-graphs on the underlying biological network

  synthetic tcga embryonic simulation sub-graph reproducibility

• Dynamic Gene Attention Focus (DyGAF): Enhancing Biomarker Identification Through Dual-Model Attention Networks.

  PMID 40160891 · PMC11951896 · Bioinformatics and biology insights · 2025 · 6 claims · 5 setups

  Has reproduction · 44

  DyGAF, a dual-model attention neural network (independent Model A + dependent Model B), identifies and ranks genes by significance for COVID-19 biomarker discovery more effectively than differential expression analysis (DEA) and random forest (RF) feature selection

  nasopharyngeal same dygaf-ranked bulk differential random

• Multi-omics feature engineering driven by biomedical foundation models improves drug response prediction for inflammatory bowel disease patients.

  PMID 41844950 · PMC13129071 · Scientific reports · 2026 · 8 claims · 7 setups

  Full-text index only

  FM (MAMMAL)-derived drug-target binding affinity (BA) inference can be used to rank/select biologically relevant protein targets and their associated genes/SNPs for a drug of interest without knowledge of protein structure or active sites

  reference top surgical snp-mutated inference gene genomic

• Gene Set Enrichment Analysis Reveals Individual Variability in Host Responses in Tuberculosis Patients.

  PMID 34421903 · PMC8375662 · Frontiers in immunology · 2021 · 8 claims · 8 setups

  Has reproduction · 71

  TB patients show substantial individual variability in the intensity of hallmark IFN responses, as well as in complement system, metabolic, and other pathway responses.

  blood cynomolgus external gene differential principal

• Deep learning based protocol to construct an immune-related gene network of host-pathogen interactions in plants.

  PMID 36525344 · PMC9791427 · STAR protocols · 2023 · 7 claims · 6 setups

  Has reproduction · 96

  DLNet algorithm ranks genes based on their contribution to classifying treatment vs. control expression data

  oryza processed rice constructed affymetrix protein-protein gene

• scLong: a billion-parameter foundation model for capturing long-range gene context in single-cell transcriptomics.

  PMID 41639087 · PMC12982784 · Nature communications · 2026 · 7 claims · 4 setups

  Full-text index only

  scLong performs self-attention across all ~27,874 human genes, including lowly expressed ones, to capture long-range gene dependencies missed by models restricted to highly expressed gene subsets

  million norman self-supervised transcriptional genetic

• Geometry-aware graph attention networks to explain single-cell chromatin states and gene expression with SEAGALL.

  PMID 42026624 · PMC13238118 · Genome biology · 2026 · 8 claims · 6 setups

  Full-text index only

  SEAGALL combines a geometry-regularised autoencoder (GRAE) to embed cells and build a cell-cell graph with a graph attention network (GAT) classifier and GNNExplainer-based XAI to identify features driving cell type/phenotype.

  count same scchip-seq ten count latent classification

• Extensive androgen receptor enhancer heterogeneity in primary prostate cancers underlies transcriptional diversity and metastatic potential.

  PMID 36450752 · PMC9712620 · Nature communications · 2022 · 8 claims · 8 setups

  Has reproduction · 70

  AR chromatin binding is highly heterogeneous between primary prostate tumors, with <5% of all AR binding sites (ARBS) shared by half of the 88 tumors analyzed

  prostate normal pca lncap chip-seq starr-seq luciferase

• Developing prognostic gene panel of survival time in lung adenocarcinoma patients using machine learning.

  PMID 35117753 · PMC8799101 · Translational cancer research · 2020 · 8 claims · 5 setups

  Has reproduction · 32

  Naïve Bayes using a 22-gene panel is the best-performing and most stable machine learning model for predicting LUAD survival time (>3 vs <3 years)

  tcga geo rna-seq methylation sequencing

• Multiple whole genome alignments and novel biomedical applications at the VISTA portal.

  PMID 17488840 · PMC1933192 · Nucleic acids research · 2007 · 8 claims · 4 setups

  Full-text index only

  A novel multiple whole-genome alignment algorithm treats all genomes symmetrically, avoiding dependence on a single base/reference genome

  genome vertebrate multi-species whole-genome conservation non-synonymous