Machine Learning models for virtual screening of new schistosomicidal compounds

Schistosomiasis is a neglected tropical disease caused by parasites from Schistosoma genus. Praziquantel is the only current treatment available. Though it is effective, it is not active against all life stages of the parasite and resistance has been reported. These limitations means there is the urgent need for the development of new schistosomicidal drugs.1 Two data sets of compounds phenotypically tested against adult and schistosomula life stages of S. mansoni were compiled from the literature. The data sets were curated according to best practices for QSAR modeling.2 Binary and continuous QSAR models were developed using fingerprint descriptors and machine learning algorithms for both data sets. The best models were used in a QSAR-based virtual screen of the ChemBridge commercial database. After clustering analysis and visual inspection, 15 compounds were selected and purchased. These compounds are under experimental validation in adult worms and juvenile life stages of S. mansoni.