Molecular versus van der Waals-like Surfaces: Revisiting The Choice Of Solute-solvent Boundary Definition In Implicit Solvent

Abstract

Implicit treatment of the solvent environment offers an optimal balance between efficiency and accuracy that greatly extends our ability to simulate protein structures and conformational transitions. The most accurate description so far is achieved by continuum dielectric solvation models, including generalized Born (GB) and Poisson-Boltzmann (PB) theories. The precise definition of the solute-solvent boundary is one of the most important features in continuum dielectric models. While it is believed that so-called molecular surfaces (MS) should provide the most physical description, most existing GB models are based on van der Waals-like (VDW) surfaces for computational simplicity and efficiency. VDW surfaces do not capture so-called reentrant surface. While it has been pointed out that VDW surface definition leads to small, solvent-inaccessible (and thus unphysical) high dielectric pockets in large proteins, the precise consequences of using VDW surfaces in simulation of smaller peptides are not well understood. In particular, it is believed by many that one might be able to compensate for drawbacks of VDW surfaces through optimization of certain parameters such as intrinsic radii of atoms. Here, we first demonstrate that such optimization has limited capability to compensate for systematic errors of VDW surfaces, which is particularly problematic for describing charged side chains and has important implications in conformational equilibrium of even small peptides. We then describe an efficient approximation of MS within the frame work the generalized Born with a simple switching (GBSW) model. The new model is as efficient as the original VDW surface based GBSW model, but is able to reproduce the Born radii calculated from the MS PB theory with a correlation of 0.98. Preliminary results of optimization of the new model on peptide simulations will also be discussed.