Integration of Machine Learning‐QSAR and Polypharmacology in Drug Discovery for Leishmaniasis Using Natural and Repurposed Compounds

Abstract Leishmaniasis is a neglected tropical disease impacting over 1 billion people worldwide. Existing treatments have limitations in toxicity and efficacy, prompting the need for new therapies. This study combines machine learning (ML), quantitative structure‐activity relationship (QSAR), and polypharmacology to identify potential drug candidates from natural sources and repurpose existing drugs. A dataset of 12,072 compounds with IC50 values for nine Leishmania species, described by PubChem fingerprints, was analyzed. Using the random forest and XGBoost models, the study achieved predictive accuracy with R 2 scores of 0.70‐0.71 and RMSE metrics around 0.46‐0.47. Key findings include natural products from worldwide biodiversity, such as 16‐acetoxy‐11‐hydroxyoctadeca‐17‐ene‐12,14‐diynyl acetate, 12‐methyl‐E,E‐2,13‐octadecadien‐1‐ol, nonadec‐5‐en‐1‐ol, and dihydrocyclobuxine D. Drug repurposing efforts highlighted significant antileishmanial activity with nystatin, stamycin, spiramycin, solamargine, and plicamycin. This approach highlights ML's potential in accelerating leishmaniasis treatment discovery.