High-Throughput Prediction of the Hydration Free Energies of Small Molecules from a Classical Density Functional Theory Yu Liu, Jia Fu, and Jianzhong Wu* Departments of Chemical and Environmental Engineering and Mathematics, University of California, Riverside, California 92521, United States * S Supporting Information ABSTRACT: The classical density functional theory (DFT) is proposed as an efficient computational tool for high-throughput prediction of the solvation free energies of small molecules in liquid water under the ambient condition. With the solute molecules represented by the AMBER force field and the TIP3P model for the solvent, the new theoretical method predicts the hydration free energies of 500 neutral molecules with average unsigned errors of 0.96 and 1.04 kcal/mol in comparison with the experimental and simulation data, respectively. The DFT predictions are orders of magnitude faster than conventional molecular dynamics simulations, and the theoretical performance can be further improved by taking into account the molecular flexibility of large solutes. SECTION: Liquids; Chemical and Dynamical Processes in Solution S olvation free energy plays a key role in solution chemistry, and its theoretical prediction often represents a bottleneck in understanding important chemical and biological processes in water including protein-ligand bindings. 1,2 The solvent- solute interactions entail complex microscopic details that make a reliable prediction of solvation properties a serious computa- tional challenge. The conventional methods for solvation free- energy calculations are based on either the solvent-implicit models or the solvent-explicit models. The former are usually constructed from knowledge-based macroscopic considerations, while the latter often start with a semiempirical force field to represent the solvent-solvent and solvent-solute interactions. 3 The continuous and solvent-explicit methods are complemen- tary, and their choices for practical applications often reflect a compromise of the computational cost and the precision in microscopic details. With the solvent molecules depicted as a dielectric continuum, an implicit-solvent model describes the solvation- free energy in terms of the geometric measures of the solute- solvent boundary, such as the solvent-accessible surface (SAS), and various energetic contributions due to the solute-solvent electrostatic and van der Waals interactions. 4 A recent example for the implicit-solvent approach was provided by Boyer and Bryan. 5 who correlated the hydration free energy with the SAS and the partial charges of the solute molecules. Impressive fitting of experimental results was achieved with a mean absolute error of 0.513 kcal/mol for the hydration free energies of diverse chemical species. While application of a continuous model drastically oversimplifies solvent-solute interactions and thus reduces the computational cost for predicting the solvation free energy and solvent-mediated interactions, it neglects the local solvent inhomogeneity and the steric effects affiliated with individual solvent molecules. Recently, Nakamura et al. 6 introduced an elegant procedure to account for the local dielectric inhomogeneity near the solute using field-theoretic techniques. The new theoretical method predicts ionic solvation free energies in both single-component liquids and binary liquid mixtures, in excellent agreement with the experimental data. The general applicability of this method to more complicated chemical systems is yet to be established. The explicit-solvent models, which have been widely used in molecular simulation and theoretical investigation of solvation, provide a more realistic representation of the solute-solvent interactions on the length scale pertinent to the solvent-solute interactions. Recent years have witnessed rapid progresses in the development of novel theoretical and simulation methods for efficient predictions of solvation free energies with explicit solvents. Analytical predictions of solvation free energies are mostly based on the integral equation theories 7,8 or the classical density functional methods. 9-15 In particular, the molecular density functional theory (MDFT) of solvation recently proposed by Borgis and coworkers offers an excellent platform to study solvation and solvent-mediated interactions. 16-18 Like the molecular Ornstein-Zernike (MOZ) equation, the MDFT is mainly applicable to rigid molecules and involves multi- dimensional density profiles that are numerically inconvenient. There have been several reports on large-scale simulation of the solvation free energies. One of the early examples was given by Received: August 20, 2013 Accepted: October 16, 2013 Published: October 16, 2013 Letter pubs.acs.org/JPCL © 2013 American Chemical Society 3687 dx.doi.org/10.1021/jz401787p | J. Phys. Chem. Lett. 2013, 4, 3687-3691