Advancing Quantitative Structure Activity Relationship Strategies in Ligand-Based Computer-Aided Drug Design

Abstract

Quantitative structure activity relationship (QSAR) modeling using high-throughput screening (HTS) data enables the development of predictive models for in silico screening. A cheminformatics framework termed BCL::ChemInfo was developed to establish QSAR modeling for application in drug discovery. Its prediction performance was evaluated through an extensive benchmark study assessing curated datasets from PubChem. BCL::ChemInfo was applied to identify novel pathway specific inhibitors for β-hematin crystallization in Plasmodium falciparum associated with Malaria. The resulting models achieved an experimental enrichment of 44 fold compared to the initial HTS hit rate of 0.37% for compounds based on a concentration threshold of 70µM. Sampled from these identified hit compounds, 15 out of 17 molecules were confirmed to perturb the hemozoin formation pathway in P. falciparum. Another research study involved the identification of novel specific allosteric modulators for mGlu5 acting on a distinct site related to CPPHA binding. From a compound library of over four million commercially available compounds five compounds where identified through in silico screening and experimentally validated to bind exclusively to this novel site. BCL::ChemInfo was also adapted to predict small molecule properties such as the water–octanol partition coefficient (LogP). The resulting prediction accuracy surpassed the current gold standard method XLogP.