Predicting Antileishmanial Activity of Plant-Derived Compounds Using Random Forest Modeling

Leishmaniasis is a parasitic disease for which existing treatments can be toxic and costly, and resistance to current therapies is becoming increasingly common. This study implemented a systematic review of data regarding antileishmanial activity in plant-derived compounds and used machine learning techniques to train and test a Random Forest algorithm to predict the antileishmanial activity of plant-derived compounds. Asteraceae, Euphorbiaceae, Lamiaceae, and Myrtaceae plant families were identified for their high or moderate antileishmanial activity and nativity to areas of high leishmaniasis prevalence. The Random Forest model had 89% prediction accuracy, with an out-of-bag error rate of 16%.