Think Globally, Move Locally: Coarse Graining of Effective Free Energy Surfaces

摘要

We present a multi-scale simulation methodology, based on data-mining tools for the extraction of low-dimensional reduction coordinates, to explore dynamically a protein model on its underlying effective folding free energy landscape. In practice, the averaged coarse-grained description of the local protein dynamics is extracted in terms of a few reduction coordinates from multiple, relatively short molecular dynamics trajectories. By exploiting the information collected from the fast relaxation dynamics of the system, the reduction coordinates are extrapolated "backward-in-time" to map globally the underlying low-dimensional free energy landscape. We demonstrate that the proposed method correctly identifies the transition state region on the reconstructed two-dimensional free energy surface of a model protein folding transition.