High-Dimensional Parameter Search Method to Determine Force Field Mixing Terms in Molecular Simulations.

Area of Science:

• Computational chemistry
• Biophysics
• Materials science

Background:

• Molecular dynamics (MD) force fields for lipids and ions are often developed separately.
• Standard mixing rules (e.g., Lorentz-Berthelot) for Lennard-Jones interactions in MD simulations underestimate lipid-ion binding energies, leading to inaccuracies.

Purpose of the Study:

• To derive accurate explicit lipid-ion cross terms for MD simulations.
• To improve the reliability of MD simulations involving lipid bilayers and ions.

Main Methods:

• Developed explicit lipid-ion cross terms, potentially including many-body effects.
• Optimized cross terms simultaneously using high-dimensional searches with ParOpt software.
• Applied optimized cross terms in MD simulations of lipid bilayers.

Main Results:

• Reduced errors from mixing rules to below 10 kJ/mol.
• Resolved structural discrepancies between previous simulations and scattering experiments (X-ray and neutron).
• Achieved accurate simulations of lipid bilayers with ions without explicit polarization terms.

Conclusions:

• Optimized lipid-ion cross terms significantly enhance the accuracy of MD simulations.
• The developed approach can be applied to various simulations employing mixed force fields.
• While effective, the NB-fix cross terms lack a physical basis, highlighting the continued need for polarizable lipid models.