Rational Design of Nonbonded Point Charge Models for Monovalent Ions with Lennard-Jones 12-6 Potential

Metallicious: Automated Force-Field Parameterization of Covalently Bound Metals for Supramolecular Structures

Metal ions play a central, functional, and structural role in many molecular structures, from small catalysts to metal-organic frameworks (MOFs) and proteins. Computational studies of these systems typically employ classical or quantum mechanical approaches or a combination of both. Among classical models, only the covalent metal model reproduces both geometries and charge transfer effects but requires time-consuming parameterization, especially for supramolecular systems containing repetitive units. To streamline this process, we introduce metallicious, a Python tool designed for efficient force-field parameterization of supramolecular structures. Metallicious has been tested on diverse systems including supramolecular cages, knots, and MOFs. Our benchmarks demonstrate that parameters accurately reproduce the reference properties obtained from quantum calculations and crystal structures. Molecular dynamics simulations of the generated structures consistently yield stable simulations in explicit solvent, in contrast to similar simulations performed with nonbonded and cationic dummy models. Overall, metallicious facilitates the atomistic modeling of supramolecular systems, key for understanding their dynamic properties and host-guest interactions. The tool is freely available on GitHub (https://github.com/duartegroup/metallicious).

Simulating Metal-Imidazole Complexes

One commonly observed binding motif in metalloproteins involves the interaction between a metal ion and histidine's imidazole side chains. Although previous imidazole-M(II) parameters established the flexibility and reliability of the 12-6-4 Lennard-Jones (LJ)-type nonbonded model by simply tuning the ligating atom's polarizability, they have not been applied to multiple-imidazole complexes. To fill this gap, we systematically simulate multiple-imidazole complexes (ranging from one to six) for five metal ions (Co(II), Cu(II), Mn(II), Ni(II), and Zn(II)) which commonly appear in metalloproteins. Using extensive (40 ns per PMF window) sampling to assemble free energy association profiles (using OPC water and standard HID imidazole charge models from AMBER) and comparing the equilibrium distances to DFT calculations, a new set of parameters was developed to focus on energetic and geometric features of multiple-imidazole complexes. The obtained free energy profiles agree with the experimental binding free energy and DFT calculated distances. To validate our model, we show that we can close the thermodynamic cycle for metal-imidazole complexes with up to six imidazole molecules in the first solvation shell. Given the success in closing the thermodynamic cycles, we then used the same extended sampling method for six other metal ions (Ag(I), Ca(II), Cd(II), Cu(I), Fe(II), and Mg(II)) to obtain new parameters. Since these new parameters can reproduce the one-imidazole geometry and energy accurately, we hypothesize that they will reasonably predict the binding free energy of higher-level coordination numbers. Hence, we did not extend the analysis of these ions up to six imidazole complexes. Overall, the results shed light on metal-protein interactions by emphasizing the importance of ligand-ligand interaction and metal-π-stacking within metalloproteins.

Viscosity Prediction of High-Concentration Antibody Solutions with Atomistic Simulations

The computational prediction of the viscosity of dense protein solutions is highly desirable, for example, in the early development phase of high-concentration biopharmaceutical formulations where the material needed for experimental determination is typically limited. Here, we use large-scale atomistic molecular dynamics (MD) simulations with explicit solvation to de novo predict the dynamic viscosities of solutions of a monoclonal IgG1 antibody (mAb) from the pressure fluctuations using a Green-Kubo approach. The viscosities at simulated mAb concentrations of 200 and 250 mg/mL are compared to the experimental values, which we measured with rotational rheometry. The computational viscosity of 24 mPa·s at the mAb concentration of 250 mg/mL matches the experimental value of 23 mPa·s obtained at a concentration of 213 mg/mL, indicating slightly different effective concentrations (or activities) in the MD simulations and in the experiments. This difference is assigned to a slight underestimation of the effective mAb-mAb interactions in the simulations, leading to a too loose dynamic mAb network that governs the viscosity. Taken together, this study demonstrates the feasibility of all-atom MD simulations for predicting the properties of dense mAb solutions and provides detailed microscopic insights into the underlying molecular interactions. At the same time, it also shows that there is room for further improvements and highlights challenges, such as the massive sampling required for computing collective properties of dense biomolecular solutions in the high-viscosity regime with reasonable statistical precision.

Self-Consistent Implementation of a Solvation Free Energy Framework to Predict the Salt Solubilities of Six Alkali Halides

To assess the salt solubilities of six alkali halides in aqueous systems, we proposed a thermodynamic cycle and an efficient molecular modeling methodology. The Gibbs free energy changes for vaporization, dissociation, and dissolution were calculated using the experimental data of ionic thermodynamic properties obtained from the NBS tables. Additionally, the Marcus' and Tissandier's solvation free energy data for Li+, Na+, K+, Cl-, and Br- ions were compared with the conventional solvation free energies by substituting into our self-consistent thermodynamic cycle. Furthermore, Tissandier's absolute solvation free energy data were used as the training set to refit the Lennard-Jones parameters of OPLS-AA force field for ions. To predict salt solubilities, an assumption of a pseudo-solvent was proposed to characterize the coupling work of a solute with its environment from infinitely diluted to saturated solutions, indicating that the Gibbs energy change of solvation process is a function of ionic strength. Following the self-consistency of the cycle, the newly derived formulas were used to determine the salt solubilities by interpolating the intersection of Gibbs free energy of dissolution and the zero free energy line. The refined ion parameters can also predict the structure and thermodynamic properties of aqueous electrolyte solutions, such as densities, pair correlation functions, hydration numbers, mean activity coefficients, vapor pressures, and the radial dependences of the net charge at 298.15 K and 1 bar. Our method can be used to characterize the solid-liquid equilibria of ions or charged particles in aqueous systems. Furthermore, for highly concentrated strong electrolyte systems, it is essential to introduce accurate water models and polarizable force fields.

Accurate Metal-Imidazole Interactions

Modeling the interaction between a metal ion and small molecules can provide pivotal information to bridge and close the gap between two types of simulations: metal ions in water and metal ions in metalloproteins. As previously established, the 12-6-4 Lennard-Jones (LJ)-type nonbonded model, because of its ability to account for the induced dipole effect, has been highly successful in simulating metal ion systems. Using the potential of mean force (PMF) method, the polarizability of the metal-chelating nitrogen from two types of imidazole molecules, delta nitrogen protonated (HID) and epsilon nitrogen protonated (HIE), has been parametrized against experiment for 11 metals (Ag(I), Ca(II), Cd(II), Co(II), Cu(I), Cu(II), Fe(II), Mg(II), Mn(II), Ni(II), and Zn(II)) in conjunction with three commonly used water models (TIP3P, SPC/E, and OPC). We show that the standard 12-6 and unmodified 12-6-4 models are not able to accurately model these interactions and, indeed, predict that the complex should be unstable. The resultant parameters further establish the flexibility and the reliability of the 12-6-4 LJ-type nonbonded model, which can correctly describe three-component interactions between a metal, ligand, and solvent by simply tuning the polarizability of the chelating atom. Also, the transferability of this model was tested, showing the capability of describing metal-ligand interactions in various environments.

Impacts of targeting different hydration free energy references on the development of ion potentials

Hydration free energy (HFE) as the most important solvation parameter is often targeted in ion model development, even though the reported values differ by dozens of kcal mol^-1 mainly due to the experimentally undetermined HFE of the proton ΔG°(H⁺). The choice of ΔG°(H⁺) obviously affects the hydration of single ions and the relative HFE between the ions with different (magnitude or sign) charges, and the impacts of targeted HFEs on the ion solvation and ion-ion interactions are largely unrevealed. Here we designed point charge models of K⁺, Mg²⁺, Al³⁺, and Cl^- ions targeting a variety of HFE references and then investigated the HFE influences on the simulations of dilute and concentrated ion solutions and of the salt ion pairs in gas, liquid, and solid phases. Targeting one more property of ion-water oxygen distances (IOD) leaves the ion-water binding distance invariant, while the binding strength increases with the decreasing (more negative) HFE of ions as a result of a decrease in ΔG°(H⁺) for the cation and an increase in ΔG°(H⁺) for the anion. The increase in ΔG°(H⁺) leads to strengthened cation-anion interactions and thus to close ion-ion contacts, low osmotic pressures, and small activity derivatives in concentrated ion solutions as well as too stable ion pairs of the salts in different phases. The ion diffusivity and water exchange rates around the ions are simply not HFE dependent but rather more complex. Targeting both the aqueous IOD and salt crystal properties of KCl was also attempted and the comparison between different models indicates the complexity and challenge in obtaining a balanced performance between different phases using classical force fields. Our results also support that a real ΔG°(H⁺) value of -259.8 kcal mol^-1 recommended by Hünenberger and Reif guides ion models to reproduce ion-water and ion-ion interactions reasonably at relatively low salt concentrations. Simulations of a metalloprotein show that a relatively more positive ΔG°(H⁺) for Mg²⁺ model is better for a reasonable description of the metal binding network.

Atomistic Simulation of Lysozyme in Solutions Crowded by Tetraethylene Glycol: Force Field Dependence

The behavior of biomolecules in crowded environments remains largely unknown due to the accuracy of simulation models and the limited experimental data for comparison. Here we chose a small crowder of tetraethylene glycol (PEG-4) to investigate the self-crowding of PEG-4 solutions and molecular crowding effects on the structure and diffusion of lysozyme at varied concentrations from dilute water to pure PEG-4 liquid. Two Amber-like force fields of Amber14SB and a99SB-disp were examined with TIP3P (fast diffusivity and low viscosity) and a99SB-disp (slow diffusivity and high viscosity) water models, respectively. Compared to the Amber14SB protein simulations, the a99SB-disp model yields more coordinated water and less PEG-4 molecules, less intramolecular hydrogen bonds (HBs), more protein-water HBs, and less protein-PEG HBs as well as stronger interactions and more hydrophilic and less hydrophobic contacts with solvent molecules. The a99SB-disp model offers comparable protein-solvent interactions in concentrated PEG-4 solutions to that in pure water. The PEG-4 crowding leads to a slow-down in the diffusivity of water, PEG-4, and protein, and the decline in the diffusion from atomistic simulations is close to or faster than the hard sphere model that neglects attractive interactions. Despite these differences, the overall structure of lysozyme appears to be maintained well at different PEG-4 concentrations for both force fields, except a slightly large deviation at 370 K at low concentrations with the a99SB-disp model. This is mainly attributed to the strong intramolecular interactions of the protein in the Amber14SB force field and to the large viscosity of the a99SB-disp water model. The results indicate that the protein force fields and the viscosity of crowder solutions affect the simulation of biomolecules under crowding conditions.