Topological data analysis based on mixed-solvent molecular dynamics simulations enhance cryptic pocket detection

Abstract

ABSTRACTSome functional proteins change their conformation to reveal a hidden binding site when approached by a binding molecule. Such binding sites are called cryptic sites and are important targets for drug discovery. However, it is still difficult to predict cryptic sites correctly. Therefore, we propose a new method to correctly detect cryptic sites using persistent homology method, one of the topological data analysis and mixed-solvent molecular dynamics (MSMD) simulations. To detect hotspots corresponding to cryptic sites, we performed the MSMD simulations using six different probes with different chemical properties (Benzene, Isopropanol, Phenol, Imidazole, Acetonitrile, and Ethylene glycol) and then performed our topological data analysis method, DAIS, to rank hotspots in the order of descending possibility of cryptic sites. For nine target proteins harboring cryptic sites, the proposed method significantly outperformed the accuracy of recent machine learning methods. As a result, in 6 out of 9 cases, the correct hotspots were ranked 1. The proposed method searched for hotspots on the “ligandable” protein surface using MSMD simulations with six different probes and detected potential cryptic sites based on DAIS-based estimates of the protein’s “conformational variability”. This synergistic combination allowed us to predict cryptic sites with high accuracy.