Open Force Field BespokeFit: Automating Bespoke Torsion Parametrization at Scale.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Accurate transferable force fields are crucial for atomistic modeling of biological processes like protein-ligand binding.
• Current force fields, while advanced, struggle with transferable torsion parameters due to complex effects.
• Bespoke parametrization is often required for these critical parameters.

Purpose of the Study:

• To present novel software, QCSubmit and BespokeFit, for scalable fitting of force field torsion parameters.
• To demonstrate the software's ability to generate quantum mechanical reference data and derive bespoke parameters.
• To evaluate the impact of bespoke torsion parameters on the accuracy of molecular modeling and binding free energy calculations.

Main Methods:

• Utilized Open Force Field QCSubmit to generate 671 quantum chemical torsion scans for druglike fragments.
• Employed BespokeFit to derive individual torsion parameters from the quantum mechanical reference data.
• Calculated relative binding free energies for TYK2 inhibitors using both original and bespoke force fields.

Main Results:

• Reduced root-mean-square error in potential energy surface from 1.1 to 0.4 kcal/mol using bespoke parameters.
• Improved accuracy in relative binding free energy calculations for TYK2 inhibitors (MUE reduced, R^2 correlation increased).
• Demonstrated significant enhancement in force field accuracy through scalable, automated parametrization.

Conclusions:

• QCSubmit and BespokeFit effectively facilitate the large-scale fitting of torsion parameters to quantum mechanical data.
• Bespoke torsion parameters derived using this approach significantly improve the accuracy of molecular mechanics force fields.
• This methodology enhances the reliability of atomistic modeling for applications such as drug discovery.