Exploring protein conformations by cluster-guided iterative multiple independent molecular dynamics simulations

Enhanced sampling methods in molecular dynamics (MD) simulations have been gaining popularity in the past decades because they can explore conformations of proteins more efficiently than conventional MD simulations. In this paper, we implement a protocol of enhanced sampling that combines iterative multiple independent MD simulations and cluster analysis. After a set of independent simulations, the combined trajectory is divided into clusters. The representative structures picked from the clusters are utilized to start the next cycle of MD simulations. By using different strategies to pick the representative structures, the enhanced sampling can be either targeted or non-targeted. Two multi-domain proteins, Escherichia coli adenylate kinase (AdK) and the three-domain (PHD-Bromo-PWWP) structure in the BS69 protein, were selected to test the method. The data indicate that conformations of the proteins can be efficiently explored, and the results show better agreement with the experimental data than those obtained through conventional MD simulations.