The ability of chemists to perform late-stage functionalization on a particular carbon-hydrogen (C-H) bond within a substrate is of high synthetic utility as it allows for the speedy modification of the compounds of interest and offers the ability to quickly alter their biological properties. One such reaction, a C-H borylation, furnishes substrates that can serve as valuable late-stage coupling partners in the Suzuki-Miyaura reaction. However, it is often difficult to predict whether a substrate is competent in site-selective borylations. Supervised machine learning and computational chemistry methods were used to classify potential borylation sites in small molecules as borylating or non-borylating. Several machine learning algorithms including random forest, neural network, and support vector machines were evaluated against a complied database of 255 test substrates taken from the literature. Features of the borylation reaction and computationally determined electronic characteristics of each substrate were used in the classification models. The insights gained from these studies will be presented.

Insight into the regioselectivity of C-H borylations employing classification machine learning models

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.