Jung-Hsin Lin, Academia Sinica

Abstract

Conformational sampling for biological macromolecules, e.g., proteins, DNA, RNA, etc., usually relies on molecular dynamics simulations (MDS), either in explicit solvent or with continuum electrostatic models to mimic the physiological environment of the biomolecules. MDS are intrinsically an N-particle move conformational sampling algorithm, in which the movement of each atom is guided by the force experienced. To avoid steric collision, the movement for each step has to be small so that MDS can be conducted smoothly. MDS usually will be terminated if the system evolves into bad configurations. Another popular sampling approach is the Monte Carlo method, and the Metropolis algorithm is commonly applied if the canonical ensemble distribution is desired. Monte Carlo methods have the advantages of being more robust, and can be conducted with bad initial configuration. However, there is no efficient N-particle move algorithm for the Monte Carlo sampling of biomolecular conformations, if the force calculations are to be avoided. We recently proposed a new method to construct the protein conformations by solving the discretizable distance geometry problem with interval data, and we now apply this method to generate the conformations for the Monte Carlo move of the protein conformations. We will demonstrate the efficiency of sampling of our method with various sampling strategies for folded protein structures, and also test whether our method could be suitable protein folding simulations.

Joint work with Antonio Mucherino, IRISA, University of Rennes 1, France.

Video