Solvation Model Based on Weighted Solvent Accessible Surface Area

We have developed a fast procedure to predict solvation free energies for both organic and biological molecules. This solvation model is based on weighted solvent accessible surface area (WSAS). Least-squares fittings have been applied to optimize the weights of SAS for different atom types in order to reproduce the experimental solvation free energies. Good agreement with experimental results has been obtained. For the 184-molecule set (model I), for which there are experimental solvation free energies in 1-octanol, we have achieved an average error of 0.36 kcal/mol, better than that of the SM5.42R universal solvation model1 by Li et al. For the 245-molecule set (model II) that has experimental aqueous solvation free energies, our WSAS model achieves an average error of 0.48 kcal/mol, marginally larger than that of Li's model (0.46 kcal/mol). We have used a 401-molecule set, the largest training set (model IV) that we know of solvation model development, to derive the SAS weights in order to reproduce the experimental solvation free energies in water. For this model, we have achieved an average unsigned error of 0.54 kcal/mol and an RMS error of 0.79 kcal/mol.The advantage of this model lies in its simplicity and independence of charge models. We have successfully applied this model to predict the relative binding free energies for the five binding modes of HIV-1RT/efavirenz. The most favorable binding mode, which has an RMSD of 1.1 Å (for 54 Cα around the binding site) compared to the crystal structure, has a binding free energy at least 10 kcal/mol more negative than the other binding modes. Moreover, the solvation free energies with WSAS have a high correlation (the correlation coefficient is 0.92) to the solvation free energies calculated by the Poisson−Boltzmann/surface area (PBSA) model. As an efficient and fast approach, WASA is also attractive for protein modeling and protein folding studies. We have applied this model to predict the solvation free energies of the 36-mer villin headpiece subdomain in its native structure, a compact folding intermediate, and a random coil. The rank order of the solvation free energies and the free energies for the three kinds of conformational clusters are in reasonable agreement with those found by MM-PBSA, a widely used solvation free energy model.