TRACE: Data for Cross-Disease Transcriptomic Drug Repurposing

TRACE (Transcriptomic Reversal And Convergent Evidence) is a computational drug-repurposing framework that scores 1,768 LINCS L1000 small-molecule compounds by their capacity to reverse disease-associated gene expression signatures, benchmarked across three independent diseases spanning fibrotic and autoimmune/inflammatory pathophysiology: idiopathic pulmonary fibrosis (IPF), rheumatoid arthritis (RA), and ulcerative colitis (UC).

The dataset is organized into eleven folders: differential_expression (per-cohort differential expression results for all 10 GEO cohorts used across the three diseases), disease_signatures (meta-analysis consensus signatures for IPF, RA, and UC), drug_scores (reversal and network-propagated scores for all 1,768 drugs across all three diseases), known_actives (curated positive and negative control compound lists per disease), crispr (genome-wide CRISPR knockout reversal screen results), l2s2 (expanded 2,834-compound drug-universe reversal scores), scrna (single-cell AT2→AT1 alveolar epithelial transition signature and associated drug scores), vae (variational autoencoder latent-space drug embeddings and reversal scores), faers (FDA Adverse Event Reporting System pharmacovigilance disproportionality results), mendelian_randomization (drug-target Mendelian randomization results for HMGCR against IPF), and validation (external validation results, including cross-cohort reproducibility on an independent held-out cohort, preclinical active-set enrichment, head-to-head comparison against the RGES state-of-the-art method, and permutation/bootstrap statistics).