VIRTUAL SCREENING AND MACHINE LEARNING-BASED IDENTIFICATION OF TROPICAL PLANT METABOLITES WITH MULTIPLE Mycobacterium ulcerans PROTEIN TARGETING POTENTIAL IN BURULI ULCER MANAGEMENT

Abstract

Buruli ulcer, a debilitating skin disease caused by Mycobacterium ulcerans, remains a neglected tropical disease with limited and often toxic treatment options. This study investigates the therapeutic potential of tropical medicinal plants using integrative computational strategies to identify bioactive compounds targeting key M. ulcerans proteins. Nine African plant species were analyzed via gas chromatography-mass spectrometry (GC-MS), yielding 217 phytochemicals. These were screened against four essential M. ulcerans enzymes - Cytochrome P450, Phosphopantetheinyl transferase, Dihydrofolate reductase, and Lysyl-tRNA synthetase using validated molecular docking protocols. ADME profiling evaluated pharmacokinetics and druglikeness, while Principal Component Analysis (PCA) explored the influence of physicochemical properties on target affinity. Machine learning-based AutoQSAR modeling was used to predict minimum inhibitory concentrations (MICs). Hit compounds such as 12-(acetyloxy)-17-(1-methyl- 4-oxo-5-phenylpentyl)gonan-3-yl acetate (AMPA) from Phyllanthus amarus and ergosta-5,24-dienol from Ageratum conyzoides showed strong binding affinities, outperforming native ligands. AMPA demonstrated the most potent predicted MIC (1.20?µg/mL), comparable to standard antibiotics and superior to previously reported phytochemicals. ADME and PCA analyses highlighted lipophilicity and hydrogen bonding as key factors in target engagement. This study validates the ethnopharmacological relevance of tropical plants and illustrates how AI-driven screening can accelerate the discovery of plant-derived therapies. The findings offer a foundation for in vitro validation and rational development of accessible, low-toxicity treatments for Buruli ulcer.