Machine Learning Assisted Exploration of Natural Monoamine Oxidase‐B Inhibitors Through Molecular Docking, Structural Dynamics and Pharmacokinetic Evaluation

ABSTRACT

Monoamine oxidase B (MAO‐B) enzyme involves in the catabolism of dopamine and is increasingly associated with neuronal damage and oxidative stress in progressive neurological diseases. This study applied consensus machine learning (ML), molecular docking, and molecular dynamics simulation (MDS) approach to screen 595,133 molecules obtained from the COCONUT database targeting the MAO‐B enzyme. Fifty promising molecules were identified as candidates with a prediction probability exceeding 90% in virtual screening. Molecular docking calculation identified molecules with PubChem CIDs of 4631050 (M25), 3951528 (M45), 1999754 (M42), 2023746 (M02), and 2004439 (M22) as the top five candidates having binding scores from −13.4 to −13.1 kcal/mol, exhibiting better values than that of the native ligand (−9.6 kcal/mol). The stability of ligand‐MAO‐B complexes were confirmed through triplicate MDS runs. Favourable binding free energy changes (−41.36 ± 3.30 to −26.80 ± 3.31 kcal/mol) indicated spontaneous and thermodynamically feasible interactions. Improved pharmacokinetic and toxicity profiles compared to selegiline and rasagiline were also demonstrated, supporting the safety and reliability of the identified hit molecules. This study proposes five molecules as potential hit candidates for the inhibition of the MAO‐B enzyme and for experimental verification. This integrative ML and in silico screening approach underscores an efficient, reliable, and economical strategy for drug design from bulk databases.