Protein Conformation Motion Modeling Using Sep-CMA-ES

Abstract

The problem of protein conformation motion modeling is an open problem in the structural computational biology. It is difficult to solve it using methods of molecular dynamics or quantum physics because these methods deal with time intervals of nanoseconds or microseconds, while conformation motions take time of millisecond order. In addition, these methods cannot take external forces into consideration. To deal with these problems, numerous approximated and coarse-grained methods are developed, which use ideas from geometry and motion planning. We present a new coarse-grained method of modeling the protein motion between two given conformations. The method is based on optimization of a cost function similar to the one in the Monge-Kantorovich mass transfer problem. The optimization is performed using sep-CMA-ES, which makes the running time of an iteration linear in the number of amino acids in a protein. The proposed method is compared with some of the existing methods on several molecules. It is shown that the results of the proposed method are more accurate than of the other methods.