Taking into Account the Ion-Induced Dipole Interaction in the Nonbonded Model of Ions

Metal ions exist in almost half of the proteins in the protein databank, and they serve as structural, electron-transfer, and catalytic elements in the metabolic processes of organisms. Molecular dynamics (MD) simulation is a powerful tool that provides information about biomolecular systems at the atomic level. Coupled with the growth in computing power, algorithms like the particle mesh Ewald (PME) method have become the accepted standard when dealing with long-range interactions in MD simulations. The nonbonded model of metal ions consists of an electrostatic plus 12–6 Lennard-Jones (LJ) potential and is used largely because of its speed relative to more accurate models. In previous work we found that ideal parameters do not exist that reproduce several experimental properties for M­(II) ions simultaneously using the nonbonded model coupled with the PME method due to the underestimation of metal ion-ligand interactions. Via a consideration of the nature of the nonbonded model, we proposed that the observed error largely arises from overlooking charge-induced dipole interactions. The electrostatic plus 12–6 LJ potential model works reasonably well for neutral systems but does struggle with more highly charged systems. In the present work we designed and parametrized a new nonbonded model for metal ions by adding a 1/<i>r</i>⁴ term to the 12–6 model. We call it the 12–6–4 LJ-type nonbonded model due to its mathematical construction. Parameters were determined for 16 +2 metal ions for the TIP3P, SPC/E, and TIP4P_EW water models. The final parameters reproduce the experimental hydration free energies (HFE), ion-oxygen distances (IOD) in the first solvation shell, and coordination numbers (CN) accurately for most of the metal ions investigated. Preliminary tests on MgCl₂ at different concentrations in aqueous solution and Mg²⁺–nucleic acid systems show reasonable results suggesting that the present parameters can work in mixed systems. The 12–6–4 LJ-type nonbonded model is readily adopted into standard force fields like AMBER, CHARMM, and OPLS-AA with only a modest computational overhead. The new nonbonded model does not consider charge-transfer effects explicitly and, hence, may not be suitable for the simulation of systems where charge-transfer effects play a decisive role