Improved continuum electrostatic modelling in proteins, with comparison to experiment.

Electrostatic interactions in macromolecules can be calculated with the method of finite differences applied to a continuum model. The accuracy of dielectric and counterion continuum modelling has been tested for long-range interactions by comparison with available experimental data over a range of ionic strengths. Various model parameters have been adjusted. Some have little effect, such as protein dielectric and the selection of Van der Waals radii. It is shown that the reduction in interaction due to dielectric effects is overestimated when a dielectric constant of 80 is assigned to all solvent accessible regions. Improved agreement is seen when the effects of the Kirkwood correlation sphere and dielectric saturation are included. Further support for the use of dielectric saturation arises from a correlation of solvent polarization saturation with crystallographic ordered water structure. Calculations over the medium ionic strength range indicate that requiring counterions to maintain a solvent layer places too great a restriction on their approach to the protein-solvent interface. However, counterion accessibility that coincides with the solvent accessible region gives too much interaction damping. Modelling of observed ion binding sites suggests that a counterion response which includes an ion desolvation term, obtained by difference calculation, will improve the computation of ionic strength effects. This study demonstrates that there is scope for improvement in continuum electrostatics calculations, and shows that progress is possible with the inclusion of physically realistic solvent and counterion properties at the protein surface.