Toward Building Protein Force Fields by Residue-Based Systematic Molecular Fragmentation and Neural Network

Area of Science:

• Computational Chemistry
• Biophysics
• Molecular Modeling

Background:

• Accurate force fields are essential for molecular dynamics (MD) simulations of biological systems.
• Developing protein force fields from quantum mechanical (QM) calculations is computationally expensive and complex.

Purpose of the Study:

• To develop an efficient and accurate method for constructing protein force fields from ab initio QM calculations.
• To overcome the computational cost and complexity limitations of traditional QM-based force field development.

Main Methods:

• Developed a residue-based systematic molecular fragmentation method to partition proteins into 20 amino acid dipeptide types and peptide bonds.
• Utilized neural network (NN) representations for QM reference data to parameterize fragments.
• Integrated NN-based QM corrections into a molecular mechanics (MM) force field baseline at the amino acid dipeptide level.

Main Results:

• The developed force fields accurately predict energies and forces for homogeneous and heterogeneous polypeptides.
• Performance was validated against full QM calculations for tripeptides to decapeptides.
• The method successfully combines classical MM force fields with NN-based QM corrections.

Conclusions:

• The residue-based fragmentation and NN parametrization method provides an efficient route to accurate, QM-derived protein force fields.
• This approach enhances the capability of MD simulations for complex biological systems.
• The study offers a novel strategy for ab initio QM-based force field development.