A methodology for analyzing nuclear Overhauser effect (NOE) data of interconverting microstates of a peptide has been suggested recently, which is based on pure statistical mechanical considerations. Thus, the mast stable microstates nd their populations are determined from the free energies, The success of this approach depends on the existence of a reliable potential energy function for the solvated peptide, in which the solvent is treated implicitly. Such a potential is developed here based on the stable structures of the cyclic hexapeptide cyclo(D-Pro(1)-Phe(2)-Ala(3)-Ser(4)-Phe(5)-Phe(6)) in DMSO obtained by Kessler et al. (J. Am. Chem. Soc. 1992, 114, 4805-4818) from NOE distance constraints, This study suggests that two different backbone motifs coexist in equilibrium, one with a beta I turn and the other with a beta II turn around Ser(4)-Phe(5). We have first reconfirmed these findings by a best-fit analysis applied to a large set of energy-minimized structures generated by our "local torsional deformations" (LTD) method, using the GROMOS force field with and without NOE distance restraints. However, the GROMOS energy E-GRO, which excludes solvent interactions was found inappropriate to describe this sq;stem because the lowest energy structures representing the beta I and pll motifs ,Ire similar to 15 and 5 kcal/mol above the global energy minimum, respectively. Solvent effects are taken into account through E-tot = E-GRO + Sigma A(i) sigma(j), where A(i) is the solvent accessible surface area (SASA) of atom i and sigma(i) is the atomic solvation parameter (ASP). We optimize the ASPs for DMSO by requiring that the E-tot values of beta I and err structures become the lowest globally, this is verified by an extensive application of LTD. The set of ASPs obtained-here will be refined in the next work where free energy (rather than energy) considerations will be taken into account. This salvation model, which is relatively easy to handle, requires significantly less computer time than explicit models of salvation and can readily be used in structural analysis of experimental data using GROMOS. The proposed derivation opens the way for the development of similar solvation models for peptides in other solvents, ASPs for rotein in water can be obtained by applying our methodology to surface Loops in proteins. Preliminary results for the ASPs, which are slightly different from the present values, were published in a recent Letter.