J Mol Graph Model. 2022 Apr 19;114:108199. doi: 10.1016/j.jmgm.2022.108199. Online ahead of print.
ABSTRACT
In this study, two approaches were applied to enhance the conformational search from molecular dynamics simulations to determine the transition states of a potential energy surface topology. The main focus is on the augmented dynamics using the swarm particle intelligence and Tsallis statistics molecular dynamics simulations of the phase transition from folding to unfolding state of a peptide in an explicit solvent environment. The transition between nodes is modelled as a random walk in a dynamic graph describing a set of basins in a free energy landscape and their pairwise relations. In this study, a dynamic graph neural networks approach is used to model the dynamic information of each free energy state as the graph evolves by observing the sequential information of edges, the time intervals between edges, and information flow. In addition, a multi-digraph approach is suggested to determine the discrete pathways of the conformation transitions between the states in that free energy surface. Besides, the role of water in the thermal and chemical denaturation of the protein is studied. This study supports the idea that the folding process is characterised by a reaction in water resulting in a reduction of the iceberg formation. Whereas unfolding by another reaction in which equilibrium is shifted towards creating iceberg states in water. In this study, the dipole-dipole correlations between the peptide and solvent are described based on an information-theoretic measure, such as local transfer entropy, to explain the role of waters in the folding/unfolding mechanism.
PMID:35462186 | DOI:10.1016/j.jmgm.2022.108199
