Metal ion–water–ligand interactions play myriad roles in biochemical, pharmaceutical, and medical sciences. For example, over 27% of all proteins have more than one metal binding site, according to the known PDB data up to the end of 2022. Because force field models are built on Newtonian mechanics and dynamics, they offer fast and detailed ways to simulate, analyze and visualize chemical processes at the level of atoms. The accuracy and transferability of force field models can be further improved using suitable Molecular Dynamics protocols or data analysis workflows. Nonetheless, the most significant challenge that force field models face is related to metal ions, specifically, simulating metal ion–ligand interactions in water or more complex systems such as metalloproteins. Various force field models have been developed and are in high demand. These force field models include the bonded, nonbonded, bonded non-bonded hybrid models featuring cationic dummy atoms and polarizable models featuring Drude oscillators with fluctuating charges. In the past six years, the work covered by this dissertation was mainly focused on providing parameters and improving the performance of the nonbonded models, which are the most widely used force field models. First, it was discovered that the traditional 12-6 Lennard - Jones model has limited accuracy when describing metal ion–water systems. This makes it hard to reproduce different experimental properties simultaneously. Based on the physical origins of the 12-6 model, an augmented nonbonded model, named the 12-6-4 LJ-type nonbonded model, has been previously proposed. However, this 12-6-4 LJ-type nonbonded model was not compatible with various new water models, like OPC3, OPC, TIP3P-FB and TIP4P-FB. Therefore, 60 different ions (including 8 monovalent cations, 4 monovalent anions, 24 divalent cations, 18 trivalent cations, and 6 tetravalent cations) were re-parametrized for the four new water models to expand the compatibility of this 12-6-4 LJ-type nonbonded model with the new water models. Further testing on these parameters proved that the 12-6-4 model could simultaneously reproduce hydration-free energies and ion-oxygen distances with high confidence levels. Second, the parameters obtained were used in the investigation of ion diffusion and ligand exchange in water. Fifteen ions (3 monovalent anions, 4 monovalent cations, 5 divalent cations, and 3 trivalent cations) had their diffusion coefficients calculated in OPC3, OPC, TIP3P-FB, and TIP4P-FB water models using an automated workflow ISAIAH (Ion Simulation using AMBER for dIffusion Action when Hydrated). A total of 60 simulated diffusion coefficients were obtained, with values within ±20% of the experimental values. Third, a pure programming contribution was needed by the AMBER molecular dynamics software package community. To improve the user-friendliness of implementing parameters for the 12-6-4 nonbonded model, modifications of the AMBER22 source code were conducted. These modifications allow users to add 12-6-4 potentials not only between two designated atom types but also between two single designated atoms, leading to specific interactions. Fourth, with proper computational support from the AMBER source code about the atom-specific pairwise potentials, the 12-6-4 potentials were then transferred to ligand (imidazole and acetate) systems. To guarantee the accuracy of ion-ligand-water simulation, the polarizability of ligating atoms was further parametrized to keep both ion-water and ion-ligand interactions consistent with the experiment. Fifth, the parameters designed for the ion-ligand-water system are applied to two protein systems. The first is an artificial protein TriCyt3 (PDB ID: 6WZC), which is used to test the robustness of the parametrization process for a protein that is totally unknown in evolution or homology models. The second is a membrane protein hZIP8 (PDB ID: 5TSA with a slight homology modification), which is used to test the transferability of these parameters in a complex environment that contains water, lipid, and protein with a variety of dielectric constants in the background.
Read
