Solvation free Energy Predictions from molecular Dynamics Simulations by improved alchemical Pathways and optimized Force Field Parameters
The rapid development of affordable medications relies on the knowledge of suitable solvents for potential active pharmaceutical ingredients. Molecular simulations can be interpreted as computational experiments and may complement laboratory experiments, as they not only enable the calculation of thermophysical properties but also allow for an insight into a systems behavior on the molecular level. Relative solubilities can be predicted by using molecular simulations to calculate the Gibbs free energy of solvation ΔGsolv. We developed algorithms to reduce the required computational effort and increase the statistical precision by improved alchemical pathways. However, the agreement between ΔGsolv predictions from simulations and experimental data depends on the molecular models (force fields) used. To consider polarization effects, we combined partial charges derived by the IPolQ-Mod method with the General Amber Force Field (GAFF) and found a comparable ΔGsolv accuracy to GAFF and its default RESP charges for a large set of compounds in various solvents. We initiated a parameter optimization to improve the accuracy of GAFF/IPolQ-Mod and present our current results of the refitting process.