We investigate the solvent-accessible area method by means of Metropolis simulations of the brain peptide Met-Enkephalin at 300 K. For the energy function ECEPP/2 nine atomic solvation parameter (ASP) sets are studied. The simulations are compared with one another, with simulations with a distance dependent electrostatic permittivity $\epsilon \mathrm{}\left(r\right),$ and with vacuum simulations $(\epsilon =2).\mathrm{}$ Parallel tempering and the biased Metropolis techniques ${\mathrm{RM}}_1$ are employed and their performance is evaluated. The measured observables include energy and dihedral probability densities, integrated autocorrelation times, and acceptance rates. Two of the ASP sets turn out to be unsuitable for these simulations. For all other systems selected configurations are minimized in the search for global energy minima, which are found for vacuum and the $\epsilon \mathrm{}\left(r\right)\mathrm{}$ system, but for none of the ASP models. Other observables show a remarkable dependence on the ASPs. In particular, we find three ASP sets for which the autocorrelations at 300 K are considerably smaller than those for vacuum simulations.

• Received 17 June 2003

DOI:https://doi.org/10.1103/PhysRevE.69.026703