Selection of Potential Ligands for TRPM8 Using Deep Neural Networks and Intermolecular Docking by the "AUTODOCK" Software

Abstract

The article describes a strategy of ligands prediction for TRPM8, where a deep neural network is used to screen out ligands and reduce the list of candidate ligands, the remaining ones are checked via AutoDock software. Subsequent analysis of the minimum binding energy between the receptor site and the putative ligands, as well as possible reactive conformations. The docking control sites were: Y745 (tyrosine 745), a critical site for TRPM8. We also analyzed the intermolecular docking of TRPM8 with its sites of manifestation of the biological effect: R1008 (phenylalanine 1008) and L1009 (alanine 1009). About 10 potential ligands were predicted, which were further verified by the "AUTODOCK" method. Intermolecular docking, carried out using the AUTODOCK program, was carried out in coordinates for each of the sites set in the closest position to the docking point. The program identified the potential for successful interactions for eight out of ten predicted candidates for each of the sites. Two of the predicted ligands do not have the ability to successfully interact with TRPM8, the rest showed a high minimum binding energy and the number of reactive conformations compared to the classical ligand, menthol. In this work, we used the method of in silico selection of ligands using deep neural network, with further verification by the AUTODOCK program. This method will speed up the search for potential medicinal substances in the future.