A strategic solution to optimize molecular docking simulations using Fully-Flexible Receptor models

• Published in: Expert systems with applications 2014-11, Vol.41 (16), p.7608-7620
• Main Authors: Vahl Quevedo, Christian, De Paris, Renata, D. Ruiz, Duncan, Norberto de Souza, Osmar
• Format: Article
• Language: English
• Subjects: Accuracy; Biological and medical sciences; Clustering of docking poses; Computation; Computer simulation; Docking; Environment models; Fully-Flexible Receptor model; Fundamental and applied biological sciences. Psychology; Interactions. Associations; Intermolecular phenomena; Ligands; Mathematical models; Molecular biophysics; Molecular docking; Receptors; Reduced Fully-Flexible Receptor model
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2014.05.038

•We optimize the molecular docking simulations of Fully-Flexible Receptor models.•We develop a heuristic method to cluster the snapshots from a molecular dynamic simulation.•We create Reduced Fully-Flexible Receptor models keeping the quality of their original models.•We select a set of representative binding poses from docking results.•We show a case of study of the Mycobacterium tuberculosis with two inhibitors. Molecular docking simulations are commonly used to identify and optimize drug candidates by examining the interactions between the target protein and small chemical ligands. This procedure is computationally expensive, especially when the receptor is treated as an ensemble of molecular dynamic conformations, namely the Fully-Flexible Receptor (FFR) model. An FFR model can vary from thousands to millions of conformations. Handling molecular docking experiments on FFR models with flexible ligands still constitutes a big challenge, since it may take hours, days, or even months to be completely executed for a single ligand. Moreover, thousands of molecular docking results are quite hard to be analyzed by a domain expert, who typically explores results starting with FEB and RMSD values. This paper addresses the high computational demand to exhaustively execute molecular docking simulations on FFR models, as well as the problem of accurately selecting a small set of representative docking results to be analyzed by a domain expert. Our approach is twofold: (1) we make use of the wFReDoW environment to decrease the dimension of the FFR model during docking experiments, trying to maintain the quality in the resulting reduced models, and (2) we perform careful analyses on docking results to select a set of representative candidate poses. Our simulation results show that the proposed method is able to achieve a trade-off between accuracy and computational cost. This is evidenced from the accuracy in wFReDoW results, which contain 96% of the snapshots within the set of the 100 best FEB values when only 67% of snapshots from the FFR model were docked.