Data-Driven Mapping of Gas-Phase Quantum Calculations to General Force Field Lennard-Jones Parameters

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Biophysics

Background:

• Molecular dynamics (MD) simulations are crucial for drug discovery and protein structure determination.
• The accuracy of MD simulations heavily relies on the chosen force field, particularly nonbonded parameters governing interactions like protein-ligand binding.
• Current force fields use atom types, leading to a large number of parameters and complex optimization challenges.

Purpose of the Study:

• To develop a novel approach for MD simulations that simplifies force field parametrization.
• To reduce the number of parameters requiring optimization in force fields.
• To improve the efficiency and accuracy of MD simulations.

Main Methods:

• Utilized the Minimal Basis Iterative Stockholder (MBIS) method for atoms-in-molecules partitioning of electron densities.
• Mapped atomic density information to Lennard-Jones parameters using a reduced set of mapping parameters.
• Eliminated the need for distinct atom types by assigning unique Lennard-Jones parameters to each atom.

Main Results:

• The new method significantly reduces the number of parameters to be optimized compared to traditional force fields.
• Achieved results comparable to the established GAFF 1.7 force field.
• Demonstrated effectiveness even when trained on limited experimental data.

Conclusions:

• The MBIS-based approach offers a more efficient and potentially more accurate method for developing molecular dynamics force fields.
• Eliminating atom types and reducing parameter optimization simplifies the parametrization process.
• This method holds promise for advancing applications in drug discovery and biomolecular simulations.