Use of osmotic coefficient measurements to validate and to correct the interaction thermodynamics of amino acids in molecular dynamics simulations

Molecular dynamics simulations are an increasingly valuable tool to biochemical researchers: advances in computational power have expanded the range of biomolecules that can be simulated, and parameters describing these interactions are increasingly accurate. Despite substantial progress in force field parameterization, recent simulations of protein molecules using state-of-the-art, fixed-charge force fields revealed that the interactions among and within protein molecules can be too favorable, resulting in unrealistic aggregation or structural collapse of the proteins being simulated. To understand why these protein-protein interactions are so over-stabilized, I first assessed the ability of simulation force fields to represent accurately the interactions of individual amino acids, employing an osmotic pressure simulation apparatus that enabled direct comparison with experiment. Surprisingly, simulations of most of the amino acids resulted in behavior that was in strong agreement with experiment. A number of amino acids, however—notably those that contain hydroxyl groups and those that carry a formal charge—interacted in ways that were clearly inaccurate. Additionally, some commonly-used force fields failed to accurately represent the interactions of amino acids in a consistent manner. By further investigating the interactions of the functional groups of these amino acids, I was able not only to determine some of the root causes of individual amino acid inaccuracies, but also to implement simple modifications that brought the interactions of these small molecules and amino acids in stronger accord with experiment. These studies have highlighted some of the shortcomings in popular simulation force fields, and have proposed useful modifications to address them. Still, there is additional work that must be—and is being—conducted in order to correctly model the interaction behavior of proteins in simulation.