Machine-Learning-Guided Discovery of Electrochemical Reactions

Area of Science:

• Organic Chemistry
• Computational Chemistry
• Electrochemistry

Background:

• Molecular structure dictates molecular function, making new reaction discovery crucial for advancing organic chemistry.
• Developing new chemical reactions provides access to novel molecules for diverse applications.
• Emerging areas in organic chemistry require efficient methods for exploring new reaction pathways.

Purpose of the Study:

• To develop a machine-learning-guided workflow for accelerating the discovery of new chemical reactions.
• To predict the competency of electrochemical reactions using machine learning models.
• To enable rapid exploration of new chemical space and facilitate the development of novel molecular functions.

Main Methods:

• Developed a novel molecular representation for creating general machine learning models with limited data.
• Utilized automated experimentation to test a large set of electrochemical reactions.
• Trained a classification model to predict reaction competency based on experimental data.

Main Results:

• Successfully screened 38,865 potential electrochemical reactions in silico.
• Identified numerous reactions of synthetic and mechanistic interest, with 80% confirmed as competent.
• Provided predictions for the entire 38,865-member set to aid future research.

Conclusions:

• A machine-learning-guided workflow can significantly expedite the discovery of competent electrochemical reactions.
• This approach enables rapid exploration of chemical space, accelerating advancements in organic chemistry.
• The developed methodology and predictions can foster faster development across various chemical disciplines.