(743g) Protein Folding Simulations with Force Fields Optimized for Secondary Structure Balance and Unfolded State Behavior

In addition to the well recognized secondary structure bias (alpha/beta balance) in all-atom protein energy functions, some other problems in the unfolded state representation are less appreciated. For example, the unfolded state is often too collapsed, resulting in a small radius of gyration compared to experiment, and the population of structured residues in the unfolded state is too high. One possible reason for the observed unfolded state properties may be that most of the force fields are parameterized using relatively primitive water models that cannot reproduce the dependence of hydrophobic hydration on temperature or pressure even for simple solutes.

In this talk, I will discuss several results from the combination of an Amber ff03-based force field, Amber ff03w, with the TIP4P/2005 water model. This new force field is parametrized, with a small backbone correction, to match the helix-coil transition of a 15-residue helical peptide to experiment. Remarkably, we find that the helix-coil transition with Amber03w is more cooperative than previous models and reproduces the T-dependence of carbonyl carbon chemical shifts accurately. The radius of gyration for nonhelical conformations is significantly larger for Amber03w and shows a collapse with increasing temperature as found in single-molecule experiments on longer proteins. A thermodynamic model is used to interpret these trends and a favorable enthalpy of solvation (higher peptide-water hydrogen bonds for nonhelical conformations) is found in case of Amber03w. The secondary structure balance is verified by the folding simulations of GB1 hairpin and Trp cage starting from completely unfolded states. In both cases we observe folded structures with backbone root mean square deviation close to 0.5 angstrom from the native PDB structures.