Ensemble-based molecular docking approach for drug resistance mutations and consensus-based reverse docking for target fishing

Abstract

Motivation: Molecular docking is a widely used technique for predicting binding interactions between a protein and a ligand at the atomic level based on protein-ligand structural information identified through X-ray crystallography or NMR. Nonetheless, it has been known that the accuracy of docking is not high. Because firstly, during the energy minimization process in docking, the ligand structure is allowed to be flexible but the receptor is not in most docking software. Secondly, the predicted ligand conformations are evaluated only once in most docking programs resulting in scoring bias. Ultimately, those two obstacles hinder the direction to the global minimum state that corresponds to the presumptive near-native binding state we are looking for.

Results: The first study is about introducing a mutation of interest to the target protein, generating ensemble homology structures, performing docking to ensembles, and identifying the mutated residues related to drug resistance. The generation of ensemble homology structures of both wild and mutant types enables the researchers to conduct flexible docking and estimate the mutational effects on drug binding sites induced by a certain mutation. The changes in docking scores and docking conformations allowed us to make mutant structure modeling by docking scores. The automated web tool is available at http://pbil.kaist.ac.kr/KRDS. The second study is about employing the strategy of reverse docking using consensus scores to reduce false positives provoked by scoring bias. The validation results showed that the consensus scoring scheme was more efficient to retrieve the known target proteins than the three existing single scoring functions when tested on 122 ligands with 6,365 known targets from the DrugBank. The web server is available at http://pbil.kaist.ac.kr/CRDS.

Conclusions: The molecular docking and reverse docking are structure-based computational schemes that can be used in various phases of drug discovery and development pipeline. During my thesis studies, I developed ensemble docking and consensus reverse docking methods to overcome two major obstacles of docking. I implemented those strategies to identify mutation residues related to drug resistance and discover clinically relevant targets of small molecules. The developed techniques have been deployed in automated web tools that perform predictive modeling and report result. The web servers enable researchers to conduct computational analyses of structural modeling for speculating possible drug-protein interactions.