Modeling deprotonation via free energy simulations (FES), which is essential for calculating pKa,  is challenging due to the need to accurately represent inter- and intramolecular interactions, which often requires quantum mechanics, while still adequately sampling conformational space, which requires long simulations. Efficiency in FES has improved recently in part due to implementation of the indirect scheme: an advantageous thermodynamic cycle can be constructed around the free energy of interest where sampling can be done at a ``low’’ (easy to simulate) level of theory, while ``high’’ (i.e., DFT or ab initio) quality results are extracted by the two connecting legs of the cycle. We have focused our efforts on improving the estimation of free energy differences along the connecting legs, i.e.,  free energy differences between levels of theory, to thereby improve the reliable prediction of pKas. Towards this aim, we have used the Quantum Mechanical Non-Boltzmann Bennett (QM-NBB) and Quantum Mechanical Non-equilibrium Work (QM-NEW) methods to calculate the pKa of ethylthiol, and we have predicted pKas of 8.0 and 10.4 respectively. Further investigation of QM-NBB results, which appears to be the poorer performer, reveals that a discrepancy between charge definitions at different levels of theory can drastically affect local solvent arrangement and thus overlap in equilibrium FES. Our labs have also identified two other culprits for poor overlap between different levels of theory: (1) stiff degrees of freedom and (2) soft degrees of freedom. All of these challenges must be overcome before using FES to accurately and reliably predict pKas in macromolecular systems with highly perturbed microenvironments.<br/>

QM-NEW: An efficient and accurate QM/MM free energy estimator and its application to pKa predictions

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.