Abstract
The falciparum malaria is a significant life-threatening disease caused by Plasmodium falciparum a protozoan parasite transmitted by the female Anopheles mosquito. The resistance of P. falciparum parasite to a limited class of antimalarial medicine has accelerated the process of screening a novel drug for falciparum malaria. In recent years the implementation of Machine Learning (ML) approaches to build a predictive model to facilitate the target-specific drug discovery process for both infectious and noninfectious pathogen has gained significance. The availability of High-throughput Screening (HTS) anti-malarial bioassay dataset has provided an opportunity to build ML-based chemoinformatics, predictive models, using features extracted from different Feature Selection (FS) algorithms. In the present study, a combination of feature selection algorithms namely Greedy Stepwise algorithm in association with CfsSubsetEval and Principal Components Analysis (PCA) in conjunction with Ranker method was used on the HTS dataset. The dataset comprising of P. Falciparum Calcium-Dependent Protein Kinase4 (PfCDPK4) inhibitors and non-inhibitors were used to train and build four state-of-art classifiers based model for predicting inhibitors of PfCDPK4 protein from an independent test dataset accurately. The classification models were evaluated based on specific statistical measures of the Weka software tool. The J48 classifier based predictive model was found to accurately predict active anti-PfCDPK4 molecule based on better Accuracy, Recall, Precision, and Area under the Curve (AUC) values. Thus, the authors conclude that the J48-based classification model will be efficient and costeffective in screening future active anti-CDPK4 molecule against P. falciparum malaria parasite. (C) 2019 The Authors. Published by IASE.