HessFit: A Toolkit to Derive Automated Force Fields from Quantum Mechanical Information

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Biophysics

Background:

• Traditional force fields (FFs) often lack precision in capturing molecular interactions due to prioritizing generality.
• Accurate simulation of molecular dynamics (MD) requires highly precise FFs, especially for complex systems.
• Quantum mechanical (QM) calculations offer detailed molecular insights but are computationally expensive for large-scale simulations.

Purpose of the Study:

• To develop and present HessFit, an open-source toolkit for refining FFs using QM-derived parameters.
• To improve the accuracy of MD simulations by integrating QM data into transferable FFs.
• To provide a robust method for parameterizing novel compounds and challenging systems like metalloproteins.

Main Methods:

• Developed HessFit, an open-source toolkit to integrate QM-derived bonded parameters and atomic charges into existing FFs.
• Utilized QM calculations (Hessian, gradient, potential energy surface) to derive dihedral barrier terms and other FF parameters.
• Evaluated HessFit by assessing vibrational and thermodynamic properties of small organic molecules and Zn2+ metalloprotein models.

Main Results:

• HessFit parameters demonstrate competitive accuracy with established FFs like GAFF2 and OPLS for small organic molecules.
• The toolkit effectively predicts thermodynamic properties of organic molecules, aligning with experimental data.
• HessFit shows comparable feasibility to current FFs for modeling nonstandard residues in Zn proteins for MD simulations.

Conclusions:

• HessFit offers a valuable approach to enhance the accuracy of molecular dynamics simulations through QM-guided force field refinement.
• The toolkit provides a feasible and effective method for deriving or extending FF parameters for novel compounds and complex biological systems.
• HessFit represents a significant advancement in molecular modeling, enabling more precise predictions across various applications.