New methods to accelerate biomolecular simulations and their applications.

摘要

This thesis describes two new methods applicable in biomolecular simulations. The P3M Ewald/rRESPA method is developed to efficiently treat the ubiquitous long-range electrostatic interactions in molecular dynamics simulations of biomolecules with the periodic boundary condition. The electric charges in the system are distributed onto a regular lattice and the fast Fourier transform (FFT) is used to calculate the Ewald energy in the k-space. The electrostatic forces are split into a fast component and a slow component, and rRESPA is used to integrate the equations of motion according to this force splitting. The P3M Ewald/rRESPA method is found to speed up molecular dynamics simulations significantly.; Water-water hydrogen bond kinetics is studied using molecular dynamics simulations with the P3M Ewald/rRESPA method. Water-water hydrogen bonds are found to persist longer in the solvation shell of hydrophobic species than hydrogen bonds in bulk water. Water polarizability is found to induce noticeable dependence of hydrogen bond kinetics on the local environment of the hydrogen bond.; The multicanonical jump walking (MJW) method is developed to efficiently sample the rugged potential energy landscapes associated with many molecular systems. In a MJW simulation, a canonical Monte Carlo sampling is made to occasionally exchange configurations with a multicanonical Monte Carlo sampling. The exchange of configurations enables the canonical sampling to overcome high energy barriers which would otherwise fragment the configuration space into regions that may be inaccessible to the finite sampling. The MJW method is extended to address the problem of global optimization of molecular conformations, and the multicanonical jump walk annealing (MJWA) method is developed. In the MJWA method, simulated annealing with gradually decreasing sampling temperature is coupled with a multicanonical sampling. The multicanonical sampling prevents the simulated annealing from being trapped in local energy minima. The MJWA method is applied to the minimization of Lennard-Jones clusters, Morse clusters and united-atom protein models, and is shown to have superior performance than the traditional simulated annealing.