Hybrid ML/MM Simulations for Computational Enzymology: Recent Advances and Challenges

Area of Science:

• Computational chemistry
• Biomolecular simulations
• Machine learning in science

Background:

• Quantum mechanics/molecular mechanics (QM/MM) with molecular dynamics (MD) is crucial for understanding enzymatic reactions.
• High computational expense of QM/MM-MD hinders the study of complex biological systems.
• Need for efficient methods to achieve atomistic resolution in biomolecular simulations.

Purpose of the Study:

• To survey the challenges and opportunities in developing hybrid machine learning/molecular mechanics (ML/MM) frameworks.
• To highlight emerging applications of ML/MM in biosystems.
• To outline future directions for accurate and efficient ML/MM models in biomolecular simulations.

Main Methods:

• Replacing quantum calculations with machine-learned interatomic potentials within a molecular mechanics framework.
• Generating high-quality reference data for training machine learning models.
• Addressing multiscale coupling challenges in hybrid ML/MM approaches.

Main Results:

• ML/MM methods offer a promising pathway to achieve near-QM/MM accuracy.
• Substantially reduced computational costs compared to traditional QM/MM-MD.
• Demonstrated potential for applications in studying complex enzymatic processes.

Conclusions:

• ML/MM methods represent a significant advancement for computational enzymology.
• Overcoming challenges in data generation and multiscale coupling is key for broader adoption.
• Future development of transferable ML/MM models will enhance next-generation biomolecular simulation workflows.