the Horizon: Multiscale Simulation of Proteins
Computer simulations of proteins have become
a common tool for complementing structural information from experiment.
Simulations can provide detailed insight into dynamic and energetic
aspects of protein structures ranging from minor structural fluctuations
to protein folding. Furthermore, computational modeling methods may
also be used to predict protein structures in part or as a whole in
the absence of suitable experimental data. Successful applications of
such methods have followed increasingly realistic model representations,
but are limited by the relatively short time scales and/or small system
sizes that can be simulated. A new, multiscale modeling strategy will
be presented that addresses these limitations and allows the modeling
of much larger systems and over much longer time scales. In this approach
approximate low-resolution models are used for fast sampling and combined
with models in atomic detail that provide a more accurate energetic
description. The efficiency of all-atom models is increased further
with a new, recently improved continuum solvation description that replaces
the need for explicit solvent molecules. Applications of such a multiscale
modeling approach to the prediction of protein structures are discussed.
Examples are presented for the modeling of missing loops in the context
of known protein structures as well as for the prediction of complete
proteins from the community-wide structure prediction experiment CASP5.