The goal of this project is to improve the quality of molecular modeling and molecular dynamics simulations, and to apply these methods to highly pertinent problems, both from the view of biomedical relevance and methodological challenge. In principle, molecular dynamics simulations offer a detailed atomic level description of the interactions between key biomolecules of significance to human health. In practice, this methodology is limited by the short simulation times (less than 100 nanoseconds) available with current computer technology, which translates into limited conformational sampling, and by the inaccuracies in the empirical force fields used in the simulations. While a number of groups are attempting to relieve the conformational sampling bottlenecks, we are focusing on improvements to the accurracy of representations, without sacrificing computational efficiency. The long range goal is to do simulations from """"""""first principles"""""""", i.e. without resorting to empirical fitting procedures. This fiscal year we developed an efficient algorithm to evaluate interatomic electrostatics up to the level of atomic hexadecapoles, using a cartesian multipole formalism and including the possibility of induced polarizable dipoles. Work on a spherical double tensor formalism in a PME context is underway. We also developed a multigrid Poisson solver for numerically evaluating the Ewald sum in the atomic multipole context. This latter approach, while not as efficient as PME on single or few processor machines, exhibits superior scaling in massively parallel machines. We successfully modeled the epsilon exonuclease of E. Coli using limited NMR-derived distance constraints, structural information from distant homologues and state of the art simulation methods.
Showing the most recent 10 out of 52 publications
