ParetoMol Benchmark Datasets: FDA-Approved Drugs, EGFR Inhibitors, and Cross-Class ADMET Compounds

Three benchmark datasets used in the ParetoMol case studies (Yabbarov & Vargas-Hernández, 2026). fda_approved_1949.csv — 1,949 FDA-approved small-molecule drugs with SMILES and 104 ADMET-AI predicted endpoints. Used in Demonstration 1 (safety-augmented Pareto analysis). egfr_top50.csv — 50 EGFR inhibitors from ChEMBL with SMILES, pChEMBL activity values, and ADMET-AI predictions. Used in Demonstration 2 (scaffold-level AhR activation analysis). cross_class_admet.json — 92 compounds across 6 drug classes (HIV protease inhibitors, SSRIs, EGFR inhibitors, beta-blockers, statins, NSAIDs) with SMILES and ADMET-AI predictions. Used in Demonstration 3 (cross-class ADMET comparison). ADMET predictions generated by ADMET-AI (Swanson et al., 2024) via the ParetoMol Tier 2 prediction service.