The global pandemic of coronavirus disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Recent World Health Organization (WHO) statistics show 778 million reported COVID-19 cases. To date, no specific drug has been found to treat COVID-19 effectively, largely due to the emergence of variants of concerns (VOCs). The main protease of SARS-CoV-2 is a well-established drug target to control viral replication in human host. We used a group of chemotypes with experimental Mpro inhibitory properties for the development of atom-based 3D-QSAR and ligand-based 3D-pharmacophore models using advanced machine learning strategies. The established 3D QSAR model is statistically significant (R2Training set = 0.9897, Q2 (R2Test set) = 0.5017), which demonstrated the model’s strong predictive power. The 3D-QSAR model displays contour maps towards the positive and negative contribution of various functional groups based on the active and inactive ligands. On the other hand, we developed a ligand-based, three-point 3D pharmacophore model using 84 ligands (39 actives and 43 inactive) that has demonstrated statistically significant data related to the discrimination of active and inactive groups of compounds with a sensitivity of 97.4%, balanced accuracy of 63.8% and a perfect ROC-AUC of 1.0, internal validation revealed that AAD2 was the best-performing hypothesis. The three-point 3D-pharmacophore model shows fitness and alignment relative to the arrangement of atoms and groups within the active and inactive subsets. In summary, the 3D QSAR and pharmacophore models developed in this study could be used as a virtual screening tool to identify virtual hits as potential SARS-CoV-2 Mpro inhibitors.
Ray et al. (Tue,) studied this question.