Energy minimization of globular proteins with solvent effects included. Comparison of empirical solvation energy terms and explicit water treatment.

The effect of an empirical solvation energy term on energy minimization of ribonuclease T1 was established using different sets of Atomic Solvation Parameters. The results are compared to minimization in vacuo and in a 10 A water shell. The best solvent model as judged from the comparison to the crystal structure was an empirical solvation potential derived from free energies of transfer of amino-acid side-chain analogues from vapour to water. The use of this model causes, however, energy and gradient oscillations, which make it inapplicable with standard protocols of molecular dynamics simulations. The empirical solvation model which was found by other authors (von Freyberg et al., 1993, J. Mol. Biol. 233, 275-292) to give good results in the NMR structure refinement led to distortions of the ribonuclease native structure. The model based on statistical analysis of crystal structures did not perform better than minimization in vacuo.