A Multiscale Approach for Path Sampling

Abstract

Proteins often accompany conformational changes when they function in cells, and its characterization is one of the most important subjects in biophysics. The ability of theoretical and computational methods based on molecular dynamics simulations has been increasing due to the advance of hardware and algorithms, it is, however, still not feasible to simulate and clarify a biological role of conformational changes of (large) proteins. To address this time-scale problem from a different point of view, we proposed to use a statistical approach based on the Onsager-Machlup (OM) action [J. Chem. Phys. 132 (2010) 134101], where a path, not a configuration, is the most fundamental object to be studied, and its probability distribution is described by a path-integral representation using the OM action in the exponent. In the present study, we apply this formalism to a model polymer system considered by Micheletti and coworkers [J. Chem. Phys. 129 (2008) 074105], where they used the OM action to estimate the free energy landscape and the diffusion coefficient of a coarse-grained variable (end-to-end distance). We execute path sampling for the same model polymer with an Asakura-Oosawa-type interaction, and address some practical issues of path sampling when we apply it to a molecular system. We further discuss how coarse-grained variables can be used to accelerate the convergence of path sampling.