Solvation Free Energies of Drug-like Molecules via Fast Growth in an Explicit Solvent: Assessment of the AM1-BCC, RESP/HF/6–31G*, RESP-QM/MM, and ABCG2 Fixed-Charge Approaches

We assess the performance of the nonequilibrium alchemical fast-growth method in calculating water and 1-octanol solvation free energies, comparing the recently proposed ABCG2 model with other empirical and quantum mechanics (QM)-based approaches for modeling electrostatic interactions in condensed phases using fixed atomic charges. The fixed-charge protocols are tested on the challenging set of drug-like polyfunctional molecules previously used by Vassetti et al., <i>J. Chem. Theory Comput.</i> <b>2019</b>, <i>15</i>, 1983-1995, broadly spanning the chemical space and often exhibiting complex conformational landscapes. We find that the cost-effective empirical ABCG2 protocol consistently outperforms the AM1/BCC precursor model and the widely used HF/6-31G* <i>ab initio</i> charge derivation method, achieving solvation free energy accuracy comparable to an expensive QM/MM-based methodology for atomic fixed charge determination. For water-octanol transfer free energies, ABCG2 benefits from systematic error cancellation, yielding remarkable agreement with experimental data, exhibiting excellent Pearson and Kendall rank coefficients and a mean unsigned error below 1 kcal/mol, matching the performance of the costly QM/MM approach. These results suggest that the ABCG2 protocol holds great promise for the high-throughput in silico prediction of ligand-protein binding free energies in drug discovery projects.