Support Vector Machine Based Prediction of P. falciparum Proteasome Inhibitors and Development of Focused Library by Molecular Docking
The emergence and spread of Plasmodium falciparum resistance to existing antimalarials emphasize the impelling search for novel drug targets and chemotherapeutic compounds. The ubiquitin-proteasome system plays a major role in overall protein turnover, in eukaryotic cells including plasmodia. 20S β subunit is the catalytic core of this proteolytic machinery, and hence most of the inhibitors developed are being targeted towards this component. Inhibition of the proteasome is established as a promising strategy to develop novel antimalarial drugs. The present study reports identification of novel drug-like 20S proteasome inhibitors with potential activity against the 20S β subunit of P. falciparum using a combination of ligand based (Support Vector Machines) and receptor based (molecular docking) techniques. The robust learning and generalizing capability of Support Vector Machines (SVM) has been exploited to classify proteasome inhibitors and non-inhibitors, targeted towards P. falciparum 20S proteasome. SVM model has been trained using 170 molecular descriptors of 64 inhibitors and 208 putative non-inhibitors of 20S proteasome. The non-linear classifier based on Radial Basis Function (RBF) kernel yielded highest classification accuracy in comparison to the linear classifier. The best classifier had 5-fold Cross-Validation (CV) accuracy of 97% and Area Under Curve (AUC) of 0.99 reflecting good accuracy of the model. The SVM model rapidly classified compounds with potential proteasomal activity. Subsequently, molecular docking studies aided the generation of focused collection of compounds with good binding affinity towards the substrate-binding site of 20S β subunit. The novel drug-like 20S proteasome inhibitors identified in this study can be a good starting point to develop novel antimalarial drugs.