Integrated machine learning and molecular dynamics–driven multi-target virtual screening of FDA-approved drugs for drug repurposing in breast cancer
摘要
Breast cancer remains the most frequently diagnosed malignancy among women worldwide, necessitating multi-target therapeutic strategies that address its molecular complexity and resistance mechanisms. This study presents an integrated computational drug repurposing framework combining machine learning (ML)-based pre-screening, molecular docking, molecular dynamics (MD) simulations, and MM-GBSA binding free energy calculations to identify FDA-approved drugs with polypharmacological potential against five breast cancer targets: HER2, EGFR, VEGFR2, HDAC3, and CDK6. 12 ML algorithms were evaluated per target (AUC-ROC: 0.878–0.951), and SHAP analysis revealed target-specific descriptor patterns. Virtual screening of 3000 FDA-approved drugs identified 94 compounds with pan-inhibitory potential across all four primary targets, from which the top 50 were validated through AutoDock Vina docking (re-docking RMSD < 2.0 Å). Ponatinib emerged as the top-ranked computational candidate (mean: − 10.07 kcal/mol), followed by Regorafenib (− 9.64), Sorafenib (− 9.46), and Entrectinib (− 9.45), while non-oncology drugs including antrafenine, betrixaban, and maraviroc demonstrated novel multi-target binding profiles. All-atom MD simulations (500 ns, CHARMM36m/CGenFF, 310 K) confirmed stable binding poses for five lead complexes, with ligand RMSD values of 1.05–1.85 Å remaining below the 2.0 Å threshold. MM-GBSA calculations revealed a binding hierarchy concordant with docking scores (R2 = 0.92): Ponatinib–VEGFR2 (ΔGbind = − 42.38 kcal/mol) > Entrectinib–CDK6 (− 38.56) > Ponatinib–EGFR (− 33.24) > Entrectinib–HER2 (− 28.47) > Dacomitinib–HDAC3 (− 24.63 kcal/mol). Energy decomposition uncovered target class–dependent thermodynamics, with van der Waals–driven kinase binding versus desolvation-penalized HDAC3 interactions. As the present study is entirely computational, the identified compounds should be regarded as hypothesis generating leads requiring experimental validation through in vitro kinase and HDAC3 inhibition assays, cell based studies, and target engagement confirmation before any translational conclusions can be drawn.