Assessing the accuracy of physical models used in protein-folding simulations: quantitative evidence from long molecular dynamics simulations

Advances in computer hardware, software and algorithms have now made it possible to run atomistically detailed, physics-based molecular dynamics simulations of sufficient length to observe multiple instances of protein folding and unfolding within a single equilibrium trajectory. Although such studies have already begun to provide new insights into the process of protein folding, realizing the full potential of this approach will depend not only on simulation speed, but on the accuracy of the physical models ('force fields') on which such simulations are based. While experimental data are not available for comparison with all of the salient characteristics observable in long protein-folding simulations, we examine here the extent to which current force fields reproduce (and fail to reproduce) certain relevant properties for which such comparisons are possible.