Influence of multidimensionality on convergence of sampling in protein simulation

Metsugi. Shoichi

Journal of the Physical Society of Japan, 74(6), p.1865 - 1870, 2005/06

https://doi.org/10.1143/JPSJ.74.1865

We study the problem of convergence of sampling in protein simulation originating in the multidimensionality of protein's conformational space. Since several important physical quantities are given by second moments of dynamical variables, we attempt to obtain the time of simulation necessary for their sufficient convergence. We perform a molecular dynamics simulation of a protein and the subsequent principal component (PC) analysis as a function of simulation time T. As T increases, PC vectors with smaller amplitude of variations are identified and their amplitudes are equilibrated before identifying and equilibrating vectors with larger amplitude of variations. This sequential identification and equilibration mechanism makes protein simulation a useful method although it has an intrinsic multidimensional nature.