Machine Learning Predicts Regioselectivity in Pd-Catalyzed Directing Group-Assisted C-H Activation

Area of Science:

• Organic Chemistry
• Computational Chemistry
• Materials Science

Background:

• Regioselectivity in C-H activation is vital for synthesizing complex molecules.
• Palladium-catalyzed C-H activation offers efficient synthetic routes.
• Predicting regioselectivity remains a challenge in catalyst design.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting regioselectivity in palladium-catalyzed C-H activation.
• To identify the most effective ML model for this prediction task.
• To accelerate the discovery of efficient catalytic systems.

Main Methods:

• Utilized machine learning techniques, including support vector machine (SVM).
• Trained and tested models on datasets of aryl substrates undergoing C-H activation.
• Evaluated model performance using metrics like F1 score and Matthews Correlation Coefficient (MCC).

Main Results:

• The standard support vector machine (SVM) model exhibited excellent generalizability.
• Achieved a high F1 score of 0.92 and MCC of 0.93 on the test set.
• Demonstrated the efficacy of ML in predicting regioselectivity.

Conclusions:

• Machine learning, particularly SVM, significantly enhances regioselectivity prediction accuracy.
• This predictive capability facilitates the rational design of catalysts for C-H activation.
• The approach supports the development of novel pharmaceuticals and advanced materials.