Paramfit: automated optimization of force field parameters for molecular dynamics simulations

Area of Science:

• Computational Chemistry
• Molecular Dynamics
• Force Field Development

Background:

• Classical molecular dynamics (MD) force fields require accurate parameters for bond, angle, and torsion terms.
• Parameter generation typically involves fitting to quantum mechanical calculations, which can be computationally intensive.

Purpose of the Study:

• To present Paramfit, a novel program for automating and enhancing the generation of MD force field parameters.
• To enable efficient parameter derivation for diverse applications, from single molecules to comprehensive force fields.

Main Methods:

• Paramfit utilizes a hybrid genetic and simplex algorithm to optimize parameters.
• It fits parameters by matching classical properties (energies, gradients) to quantum mechanical data.
• The program can derive multiple parameters simultaneously and across multiple molecules.

Main Results:

• Paramfit significantly reduces the number of quantum calculations required compared to previous methods.
• It successfully generates parameters even with sparse structural data.
• The program has been instrumental in developing the Lipid14 force field.

Conclusions:

• Paramfit offers an automated, efficient, and versatile approach to MD force field parameter generation.
• Its novel algorithmic approach streamlines the development of accurate and reliable molecular models.