Protein folding in mode space: a collective coordinate approach to structure prediction

Does the dynamics of a protein encode its structure? Many studies have addressed the inverse of this question-how a three-dimensional structure determines its dynamics-using molecular dynamics simulation, normal mode analysis, and similar methods. Recently we have developed a molecular dynamics (MD) simulation method to impose dynamic properties on ensembles of MD trajectories in the form of restraints on structural diversity in the directions of the principal components of motion of the molecule. In the current work, we investigate if these restraints in combination with a standard MD force field are sufficient to generate native structure in disordered structural ensembles. We present simulations from a series of increasingly disordered structural ensembles obtained by thermal unfolding or randomization of the coordinates of the native structure of two src-homology 3 (SH3) domains. Native structure formation is observed under the sole action of the diversity restraint and the MD force field. We investigate the importance of accuracy of the description of native dynamics. Protein folding is a highly cooperative process. The dynamic restraints may enforce long-range cooperativity and thus speed up the folding from unstructured states. Applications of the restraints to structure refinement and structure prediction are possible.