An Accurate and Efficient, Computational Method for the Hydration Free Energy of Large and Complex Molecules

Abstract

The hydration free energy (HFE) is an essential factor to characterize various chemical processes in solvent. It is then needed to construct computational methods of the HFE. In the computational point of view, not only accuracy but efficiency is necessary when large and complex molecules such as proteins are analyzed. For example, the explicit treatment of solvent is effective for accuracy, however, the treatment for protein solutions sacrifices computational efficiency. To construct a computational method satisfying both accuracy and efficiency, we decompose the HFE into two parts called the direct and indirect parts: one depends on the solute-solvent interaction, and the other is the remaining determined by the geometry of solute. The former is computed by the all-atom treatment. The latter is computed by a hybrid method of the energy representation (ER) method and the morphometric approach (MA). The ER method provides the same accuracy as standard methods such as the thermodynamic integration and free energy perturbation methods and achieves efficient computation by the end-point calculation. The ER method has computational advantageous, however, its protocols can be reduced by the hybrid method. The MA provides thermodynamic quantities primarily governed by the geometry of solute and achieves rapid computation by the fitting function constructed by geometric measures and its corresponding coefficients. Since the indirect part of the HFE depends on the geometry of solute, the MA can be applied to the indirect part and can provide same accuracy as the ER method when the coefficients of the MA are determined by fitting to values calculated by the ER method. We show the accuracy and efficiency of the hybrid method, and demonstrate performance in the discrimination of the native structure from decoys.