Automated Force Field Developer and Optimizer Platform: Torsion Reparameterization

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• General Amber Force Field (GAFF) is widely used for protein-ligand binding simulations.
• GAFF parameters, especially torsion terms, may require custom reparameterization for accurate binding free energy calculations.
• Torsion parameters are sensitive to stereoelectronic and steric effects in drug-like molecules.

Purpose of the Study:

• To develop a user-friendly computational tool, AFFDO, for generating accurate, tailored GAFF2 torsion parameters.
• To improve the accuracy of protein-ligand binding free energy simulations through custom parameterization.
• To provide a flexible platform for optimizing force field parameters for drug-like molecules.

Main Methods:

• AFFDO platform utilizes GPU-accelerated density functional theory (DFT) calculations for reference data.
• It employs a fast gradient-based optimizer with automated differentiation to fit torsion terms.
• The tool selects critical torsions for a given ligand and generates optimized GAFF2 parameters.

Main Results:

• AFFDO successfully generated improved GAFF2 torsion parameters against quantum mechanics (QM) reference data.
• Parametrization of drug-like molecules using AFFDO led to enhanced protein-ligand relative binding free energy (RBFE) simulations.
• The tool demonstrated improved agreement with experimental RBFE values in several cases.

Conclusions:

• AFFDO provides an efficient and accurate method for developing tailored GAFF2 torsion parameters.
• The platform enhances the reliability of molecular simulations in structure-based drug discovery.
• AFFDO offers a valuable solution for overcoming limitations in general force field applicability.