Development of a Novel Pharmacophore Model Guided by the Ensemble of Waters and Small Molecule Fragments Bound to SARS-CoV-2 Main Protease.

Recent fragment-based drug design efforts have generated huge amounts of information on water and small molecule fragment binding sites on SARS-CoV-2 M and preference of the sites for various types of chemical moieties. However, this information has not been effectively utilized to develop automated tools for in silico drug discovery which are routinely used for screening large compound libraries. Utilization of this information in the development of pharmacophore models can help in bridging this gap. In this study, information on water and small molecule fragments bound to M has been utilized to develop a novel Water Pharmacophore (Waterphore) model. The Waterphore model can also implicitly represent the conformational flexibilities of binding pockets in terms of pharmacophore features. The Waterphore model derived from 173 apo- or small molecule fragment-bound structures of M has been validated by using a dataset of 68 known bioactive inhibitors and 78 crystal structure bound inhibitors of SARS-CoV-2 M . It is encouraging to note that, even though no inhibitor data has been used in developing the Waterphore model, it could successfully identify the known inhibitors from a library of decoys with a ROC-AUC of 0.81 and active hit rate (AHR) of 70 %. The Waterphore model is also general enough for potential applications for other drug targets.