ABSTRACT
The biological function of a protein can be interpreted as a physicochemical process of sequential switching motions and associated chemical reactions occurring as a response to a certain external perturbation. These switching motions, or the structural transitions between the resting and the activated states, are a consequence of the highly anharmonic dynamics on the rugged potential surface of the protein molecule [1]. Thus, understanding such anharmonic dynamics is essential for elucidating the expression and regulation of molecular functions. Molecular dynamics (MD) simulation has made signi cant contributions to studies of the anharmonic dynamics of proteins. One of the most successful models for analyzing the anharmonic dynamics in MD trajectories is based on the quasiharmonic approximation, where the distribution function is approximated up to the second-order with respect to coordinates [2,3]. The anharmonicity is highlighted by embedding the MD trajectory into a few largest-amplitude modes to depict the landscape of large-scale
collective motions [4,5]. Although the quasiharmonic picture has been successfully used for analyses of the equilibrium distribution, it is not necessarily suitable for time-domain analysis of relaxation behaviors.
