Quantitative Structure-Activity Relationship Modelling of Antileishmanial Compounds Derived from Nigerian Medicinal Plants

Leishmaniasis, though less prevalent in Nigeria than some other tropical diseases, poses an emerging risk particularly in immunocompromised populations, and the limited oral bioavailability and toxicity of existing drugs create a clear need for novel antileishmanial agents. Nigerian medicinal plants represent an underexplored source of structurally diverse bioactive scaffolds. This study applied quantitative structure-activity relationship modelling to a dataset of 62 compounds isolated from Nigerian medicinal plant sources with reported antileishmanial activity. Molecular descriptors were calculated using the PaDEL-Descriptor software package. QSAR models were developed using multiple linear regression, support vector machines, and random forest algorithms, validated by internal cross-validation and external test set prediction. Model applicability domain was defined by the Williams plot. Physicochemical descriptors including topological polar surface area, molecular weight, logP, and hydrogen bond donor count were most frequently identified as significant QSAR descriptors. The best model achieved external validation R2 of 0.82 and root mean square error of 0.41 log units using the random forest algorithm. Structural features associated with enhanced activity included presence of hydroxyl and carbonyl groups on flavonoid scaffolds, and bulky lipophilic substituents on terpenoid cores. The developed QSAR models provide a computational framework for rational design and in silico screening of novel antileishmanial candidates from Nigerian phytochemical databases. Keywords: QSAR, antileishmanial, Nigerian medicinal plants, molecular modelling, machine learning.