Integrated Machine Learning and Structure-Based Virtual Screening Identifies Novel Myeloperoxidase Inhibitors with Cardioprotective Activity

Myeloperoxidase (MPO) has emerged as a promising therapeutic target for heart failure (HF) because of its pathogenic roles in oxidative stress, inflammation, and myocardial fibrosis. However, the clinical development of MPO inhibitors remains limited, and most candidates have failed in advanced trials. Here, we report the discovery of novel MPO inhibitors through an integrated computational pipeline combining machine-learning-based virtual screening with structure-based consensus docking. This approach screened ~3.1 million compounds and yielded 25 candidates for experimental evaluation. Among these, compound 16 exhibited potent MPO inhibition (IC50 = 0.297 ± 0.097 mM) and conferred significant cardioprotection in H9c2 cells subjected to oxygen-glucose deprivation, preserving cell viability to 66.19% at 25 mM. Mechanistic investigations revealed that compound 16 reduced lactate dehydrogenase release, decreased reactive oxygen species accumulation, and suppressed hypoxia/reoxygenation-induced a-SMA upregulation in a neutrophil-conditioned medium model. In addition, molecular dynamics simulations and MM-PBSA/MM-GBSA calculations suggested stable binding to MPO, primarily through p-stacking with the heme prosthetic group and hydrophobic contacts with key active-site residues. Compound 16 showed favorable predicted permeability and low predicted hERG liability, but poor metabolic stability in rat liver microsomes (T1/2 = 1.34 min), indicating that metabolic optimization is required. Collectively, this study establishes a computationally efficient strategy for MPO inhibitor discovery and identifies compound 16 as an MPO inhibitor that protects cardiomyocytes in vitro, providing a starting point for future in vivo studies aimed at heart failure therapy.