A Framework for Multiscale Characterization of Biomolecular Dynamics

Many instances of protein dynamics of relevance for biological function are now recognized in the literature: protein exibility has been related to enzyme catalysis [98], chaperon-assisted protein folding, ligand binding [96], [28], [27], molecular recognition in the immune system, etc. A major diculty in modeling protein dynamics is the time scale disparity between fast motions, in the order of fs, and motions of biological importance, which are in the order of s to seconds. A particularly well suited method for studying protein dynamics is multidimensional NMR relaxation [82], [51]. Through dierent experiments both \fast " (ps-ns) and \slow " motions (s-ms) can be accurately characterized at atomic resolution [59]. NMR can compute a large number of relaxation rates that are related to angular correlation of particular rotatable bonds in either the backbone or the sidechains of proteins. NMR experiments, however, cannot fully characterize the mechanisms of motion and conformational changes. Also, the calculation of rates assumes models of the dynamics [112], [24] that may not be applicable. One of the main simulation techniques for studying protein dynamics is molecular dynamics (MD). MD is applicable to relatively fast motions (the ps-ns range mentioned above), and can complement quite well NMR by helping to resolve relaxation rates without assuming a model [87]