David LeMaster, Ph.D.

The ability of protein molecules to switch between differing conformational states has long been recognized as providing a crucial aspect of their biological function. Quantitative understanding of these conformational transitions is particularly significant in analyzing finely-tuned regulatory processes as well as in the development of structure-based drug design. To optimize the modeling of these transitions, experimental data can identify the points of inaccuracy in the molecular simulation force fields used, and re-optimized force fields can then provide more accurate simulations which, in turn, allow for a more robust structural interpretation of the experimental data. The primary source of inaccuracy among the widely used force fields is in their handling of electrostatic energies and forces. As was early noted by Richard Feynman, parametrizations which are optimized to predict total energies need not accurately predict the forces. Nonpolarizable force fields have been optimized to predict the total electrostatic energy by artificially elevating the fixed atomic charge values in order to match the average electric field energy density using the in vacuo dielectric value of 1.

To advance this area, we are applying two types of NMR spectroscopy experiments which are acutely sensitive to the distribution of protein conformations present (amide hydrogen exchange) and the rates at which the transitions between different conformations occur (NMR relaxation). We have shown that among protein backbone amides which are well-exposed to bulk solvent, the local conformation-dependent exchange kinetics span a billion-fold range in rates. Due to the short lifetime of the anionic intermediate formed during the exchange reaction, we have found that the local conformationally-dependent pK values of these amides are predictable across this range with correlation coefficients approaching 0.95. The resultant optimal internal dielectric value of 3 for these hydrogen exchange analyses, corresponding to the neglected electronic polarizability effect, points to the inconsistency in how standard nonpolarizable molecular force fields handle electrostatic energies and electrostatic fields. This acute sensitivity in hydrogen exchange rates to differences in conformation has enabled us to quantify the site and magnitude of error that have arisen in a long timeframe simulation of bovine pancreatic trypsin inhibitor and to measure the equilibrium constant for a conformational transition in human ubiquitin which had been a point of active contention between several MD simulation studies.

To facilitate MD simulation analysis of protein NMR relaxation data, we have introduced a quaternion-based rotational velocity rescaling protocol which can enable increased accuracy in predicting relaxation behavior directly from MD simulations without invoking the problematic conventional approach of subtracting out global molecular tumbling effects. Our joint MD-NMR analysis of the human protein FKBP12 have enabled us to quantitatively model the allosteric statistical coupling of spatially distinct conformational transitions which has heretofore not been feasible in many of the more widely studied allosteric systems due to the considerably slower kinetics of those transitions.