HessFit: Deriving Molecule-Specific Force Fields from Quantum Mechanical Hessians.

The development of accurate and transferable molecular mechanics force fields (FMs) is crucial for reliable molecular dynamics (MD) simulations. Conventional force fields such as AMBER, CHARMM, OPLS, and GROMOS prioritize transferability at the expense of maximal accuracy, which can limit their performance when modeling novel chemical moieties or metal coordination environments. To address this issue, we present HessFit, an open-source Python toolkit that refines classical force fields using quantum mechanical (QM) information. HessFit derives molecule-specific bond parameters and atomic charges directly from QM Hessians and potential energy surfaces, combining analytical extraction and fitting procedures to obtain accurate bond, angle, and torsion constants. Benchmarks on several small molecules show better reproduction of vibrational frequencies, geometries, and thermodynamic properties compared to GAFF2 and OPLS. Applications to ligand-protein complexes further demonstrate HessFit's ability to model coordination environments that are typically challenging for transferable force fields. By integrating QM-level precision with classical computational efficiency, HessFit advances automated, data-driven force field parameterization, enhancing both accuracy and reproducibility across complex molecular systems.