CHARMM-GUI Hybrid ML/MM Builder for Hybrid Machine Learning and Molecular Mechanical Modeling and Simulations.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Machine learning (ML) advances are revolutionizing molecular dynamics (MD) simulations.
• Neural network potentials (NNPs) trained on quantum mechanical (QM) data offer high accuracy for molecular descriptions.
• Hybrid ML/MM methods combine NNPs for localized regions (e.g., ligands) with classical molecular mechanics (MM) for the rest of the system.

Purpose of the Study:

• To introduce CHARMM-GUI Hybrid ML/MM Builder, a tool for automating hybrid ML/MM simulations.
• To enable near-QM accuracy for protein-ligand complexes at reduced computational expense.
• To demonstrate the utility and capabilities of the new builder with application systems.

Main Methods:

• Development of CHARMM-GUI Hybrid ML/MM Builder.
• Integration with TorchANI-AMBER and OpenMM-ML for NNP-based simulations.
• Support for MACE and ANI neural network potentials.
• Automated generation of system and input files for hybrid ML/MM modeling.

Main Results:

• Successful automation of hybrid ML/MM simulation setup for protein-ligand complexes.
• Demonstration of near-QM accuracy for ligand interactions within complexes.
• Significant reduction in computational cost compared to full QM methods.
• Generation of necessary files for simulations in solution or membrane environments.

Conclusions:

• CHARMM-GUI Hybrid ML/MM Builder facilitates efficient and accurate simulation of protein-ligand complexes.
• The hybrid approach offers a powerful strategy for drug discovery and molecular modeling.
• This tool enhances the accessibility and application of advanced ML/MM techniques in computational chemistry.