Abstract
Falcipain-3 (FP-3) is a cysteine protease of the malaria parasite Plasmodium falciparum which is a promising and attractive target enzyme for antiparasitic chemotherapy. In this study, a support vector machine (SVM) model based on fingerprint-based descriptors was developed on a dataset of 239 FP-3 inhibitors to identify the most active antimalarial compounds among the active compounds provided from similarity search. The satisfactory classification performance achieved by the SVM model shows its ability to use it as a further filter to distinguish the most active compounds. The accuracy in prediction for the training, test and external validation sets were 97.39%, 94.74% and 90.6%, respectively. Furthermore, the performance of the model was examined by plotting the receiver operating characteristic (ROC) curve, and the area under the ROC curve was 0.96 for the modeling set. The ability of a virtual screening scheme to scaffold hopping or lead hopping is known as a key ability of an effective method for virtual screening. Three diverse reference FP-3 structures were chosen as active antimalarial compounds to search the lead-like database of ZINC and retrieve the most similar compounds. Compounds having Tanimoto similarity coefficient above 0.8 were extracted for further analysis by classification model. The SVM model rendered five most active compounds and they were also analyzed by ADME/Tox and diversity measures. Maximum property-based distance between extracted compounds was found to be 0.70, which shows the importance of applying multiple diverse reference compounds in the similarity searching.
Keywords: Anti-malaria, classification, fingerprint, scaffold hopping, similarity searching, SVM.
Related Journals
Current Computer-Aided Drug Design
57
2