Despite many inherent drawbacks, implicit (continuum) solvent models have become widely adopted for molecular simulation studies. In particular, pairwise generalized Born (GB) models are frequently used due to their computational simplicity. A variety of GB models have been implemented which differ in the methods used to calculate the intrinsic Born radii. These GB models typically have a large number of empirical parameters that can greatly influence their accuracy. In this presentation, we will show a new parameter set for the GB Neck model of Mongan et al., where the parameters were fit using large data sets of peptide, protein and nucleic acid solvation data from more accurate (and much slower) methods. The result is a computationally efficient GB model that provides significantly improved accuracy as compared to our previous models. After parameter refitting, we obtain a better match to Poisson-Boltzmann solvation energies, improved secondary structure propensities as compared to experiment, and better match between simulated and NMR data on protein dynamics. Implementation of this GB model in the GPU version of Amber provides the ability to access microseconds of protein MD per day on a single consumer-level GPU. We show that the simulations provide near-quantitative reproduction of thermal melting profiles for several model peptides, are capable of folding proteins with alpha, beta and mixed secondary structure to good agreement with experiment, as well as observing folding for DNA and RNA systems.

Improved continuum solvent model for simulations of proteins and nucleic acids

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.