Improved conformational sampling through an efficient combination of mean-field simulation approaches

In this article, we show for the first time the combination of locally enhanced sampling (LES) and the generalized Born continuum solvation model. It has been shown that the reduction in barrier heights provided by LES can result in solvated simulations that are more than an order of magnitude more efficient than single-copy methods, but we note that explicit solvent introduces a strong correlation between copies, reducing the benefits of the technique. The use of continuum solvation can also accelerate conformational transitions because of reduction in friction. We describe a direct combination of the two algorithms that maintains the equivalence of the LES and non-LES global energy minima and eliminates the solvent-induced coupling of the copies. Since this approach results in a significant increase in computational requirements, we also present reasonable approximations that greatly increase efficiency. We have applied the resulting combination to simulation of conformational change in an RNA UUCG tetraloop and have shown that the combined GB + LES approach is more efficient than use of either GB or LES alone. We carried out a large number of these simulations and show in a converged manner that the rate constant for the conformational transition is significantly increased with GB + LES as compared to GB alone. In addition, we demonstrate that the combined method significantly improves the ability of LES copies to explore independent transition paths as compared to LES simulations with explicit solvation. We thus believe that this GB + LES technique may be a useful component of structure refinement and prediction studies.