Abstract
The Poisson–Boltzmann (PB) continuum solvent model shows considerable promise in providing a description of electrostatic solvation effects in biomolecules, but it can be computationally expensive to obtain converged results for large systems. Here we examine the performance of a pairwise generalized Born approximation (GB) method on multiple conformations of a small peptide, three proteins (protein A, myoglobin, and rusticyanin) and four RNA and DNA duplexes and hairpins containing 20–24 nucleotides. Charge and dielectric radii models were adapted from the CHARMM and Amber force fields. Finite difference PB calculations were carried out with the Delphi and PEP programs, and for several examples the matrix of all pairwise interaction energies was determined. In general, this parameterization of the GB model does an excellent job of reproducing the PB solvation energies for small molecules and for groups near the surface of larger molecules. There is a systematic tendency for this GB model to overestimate the effects of solvent screening (compared to PB) for pairs of buried atoms, but individual errors tend to cancel, and a good overall account of conformational energetics is obtained. A simple extension to the GB model to account for salt effects (in the linearized Debye–Hückel approximation) is proposed that does a good job of reproducing the salt dependence of the PB calculations. In many cases, it should be possible to replace PB calculations with much simpler GB models, but care needs to be taken for systems with extensive burial of charges or dipoles.