Reconstructing the Most Probable Folding Transition Path from Replica Exchange Molecular Dynamics Simulations.

Abstract

The characterization of transition pathways between long-lived states, and the identification of the corresponding transition state ensembles are useful tools in the study of rare events such as protein folding. In this work we demonstrate how the most probable transition path between metastable states can be recovered from replica exchange molecular dynamic simulation data by using the dynamic string method. The local drift vector in collective variables is determined via short continuous trajectories between replica exchanges at a given temperature, and points along the string are updated based on this drift vector to produce reaction pathways between the folded and unfolded state. The method is applied to a designed beta hairpin-forming peptide to obtain information on the folding mechanism and transition state using different sets of collective variables at various temperatures. Two main folding pathways differing in the order of events are found and discussed, and the relative free energy differences for each path estimated. Finally, the structures near the transition state are found and described.