Monte Carlo vs Molecular Dynamics for Conformational Sampling

A comparison study has been carried out to test the relative efficiency of Metropolis Monte Carlo and molecular dynamics simulations for conformational sampling. The test case that has been examined is liquid hexane. OPLS potential functions have been used with sampling of external motions and internal angle bending and torsions. The BOSS 3.6 and AMBER 4.0 programs were used to perform the internal-coordinate Monte Carlo and Cartesian molecular dynamics simulations, respectively, for samples of 267 molecules in the NPT ensemble at 25 °C and 1 atm with periodic boundary conditions. The initial configurations of the system were equilibrated with the hexane molecules in the all trans configuration. The major findings are (1) the Monte Carlo and molecular dynamics results for thermodynamic properties and conformer populations are in accord, (2) the conformer populations reach equilibrium in ca. 107 Monte Carlo configurations or 100 ps of molecular dynamics, and (3) the molecular dynamics calculations require 1.6−3.8 times more computer time to achieve the same level of convergence as the Monte Carlo simulations. The viability of internal-coordinate Monte Carlo methodology contrasts with earlier notions on the inefficiency of Cartesian Monte Carlo methods. Monte Carlo sampling in related contexts such as for side-chain torsional motion in proteins appears promising.