Toward Validated Quantum Mechanical Workflows Predicting pH-Dependent Properties: Benchmarking Protocols for...

We present work toward validated quantum-mechanical workflows predicting pH-dependent properties of flexible molecules in solution. The present contribution validates a broad set of possible workflows against a focused set of pK_a data. In particular, we test two different conformational sampling protocols, four continuum solvent models, eight atomic orbital basis sets, eight density functional approximations (DFAs), two schemes for geometry optimization and incorporation of thermal effects, and three schemes for empirical correction, a total of 3076 possible approaches. We benchmark these methods' predictions for the macroscopic (experimental) pK_a of 20 N-acylsulfonamides taken from the SAMPL7 challenge. Consistent with earlier studies, we find that workflows employing standard density functionals, triple-ζ basis sets, and linear empirical corrections provide root-mean-square deviations (RMSD) in pK_a around 0.6-1.2 log units. Our workflows provide another step toward reliable black-box physics-based prediction of pH-dependent properties.