New Interaction Parameters for Charged Amino Acid Side Chains in the GROMOS Force Field

A GROMOS force-field parameter set 54A8 is developed, which is based on the latest 54A7 set [Schmid et al. Eur.: Biophys. J. 2011, 40, 843-856] and involves a recalibration of the nonbonded interaction parameters for the charged amino acid side chains, based on ionic side chain analogs. After a thorough analysis of the available experimental data, conventional hydration free energies for the ammonium; mono-, di-, tri-, and tetramethyl-ammonium; formate; acetate; propanoate; imidazolium; and guanidinium ions are combined with a standard absolute intrinsic proton hydration free energy Delta G(hyd)(circle minus)[H-g(+)] = -1100 kJ.mol(-1) to yield absolute intrinsic single-ion hydration free energies serving as experimental target data. The raw hydration free energies calculated from atomistic simulations are affected by electrostatic and finite-size artifacts, and corrections are applied to reach methodological independence prior to comparison with these experimental values. Except for monomethyl-ammonium, ions with parameters derived directly from the 54A7 force field considerably underestimate (ammonium, formate, acetate, propanoate, guanidinium) or overestimate (di-, tri-, and tetramethyl-ammonium; imidazolium) the magnitude of the intrinsic hydration free energy, the largest deviation affecting the acetate ion (40.0 kJ.mol(-1)). After reparameterization into 54A8, the mean and maximal absolute deviations between simulated and experimental data over the set of 10 ions are reduced from 23.1 and 40.0 kJ.mol(-1), respectively, to 1.8 and 6.3 kJ.mol(-1), respectively. Although the 54A7 and 54A8 parameter sets differ significantly in terms of the hydration free energies of the ions considered, other properties such as ion-water radial distribution functions and ion-ion potentials of mean force appear to be only moderately sensitive to this change. These properties are similar for the two sets and, in the case of the ion-water radial distribution functions, in good agreement with available experimental data.