Machine Learning to Predict Diels-Alder Reaction Barriers from the Reactant State Electron Density.

Area of Science:

• Computational chemistry
• Chemical physics
• Machine learning in chemistry

Background:

• Reaction barriers are fundamental to chemical reactivity and catalysis.
• Numerous studies have computed and measured reaction barriers for seminal chemical reactions.
• Existing data presents an opportunity for machine learning (ML) prediction models.

Purpose of the Study:

• To develop a machine learning model for accurate prediction of reaction barrier energies.
• To utilize topological descriptors of quantum mechanical charge density as input features.
• To demonstrate the model's efficacy on the Diels-Alder reaction and its variants.

Main Methods:

• Employed supervised regression algorithms trained on quantum mechanical charge density topological descriptors.
• Focused on Diels-Alder reactions in solution for initial training.
• Validated the model on catalyzed Diels-Alder reactions, including those with artificial enzymes.

Main Results:

• Topological descriptors of reactant-state quantum mechanical charge density accurately predict reaction barrier energies.
• The model successfully predicted barriers for complex catalyzed Diels-Alder reactions, correlating with experimental catalytic activity (kcat).
• Achieved high accuracy in barrier energy predictions across diverse reaction contexts.

Conclusions:

• Topological descriptors offer a rigorous and continuous set for predicting reaction barriers.
• The developed ML tool can predict reaction barriers accurately, reducing the need for extensive computations or experiments.
• This approach is broadly applicable to various reactions in solution and catalyzed systems for screening and analysis.