Large-Scale Computation and Prediction of Sulfinate-Mediated C-H Functionalisation Regioselectivity
Publication Date
July 30, 2025
Creator
Abstract
Sulfinate-mediated radical C–H functionalisation reactions are widely used for the modification and diversification of scaffolds in drug discovery. However, prediction of the regiochemistry in these reactions can be challenging. For a given substrate, there may be multiple sites of reaction, each with its own unique steric and electronic environment. Here we present Rega, an automated transition state searching program for the prediction of regioselectivity from inexpensive HF/6-31G* activation energies. We show that in a set of 23 compounds, the regioselectivity is correctly identified in 22 cases (reactivity correctly identified for 65/68 potential sites of reaction). The easy-to-use and modular Rega workflow allows reaction exploration of multiple substrates simultaneously, enabling the generation of a synthetic dataset of 490 compounds consisting of 2780 sites with labelled reactivity for this reaction for use in machine learning models. Rega is designed to be readily extensible to other reaction systems and can be applied to many other reaction classes in which a radical intermediate is formed as the regiochemistry determining step. From the generation of this dataset, machine learning was applied to predict regioselectivity in both regression and classification tasks.
Item Type
ethesis
Thesis Type
PhD
Supervisors
Subjects (LC)
Associated Schools / Departments
School of Chemistry
eprints ID
80983
UoN Repository URI
Except where otherwise noted, this item's license is described as
File(s)![Thumbnail Image]()
Name
Walton, Peter, 14278442, Corrections.pdf
Type
Full-text
Description
Examined. Corrections
Size
4.93 MB
Format
Adobe PDF
Checksum (MD5)
4a9ee792c29d7ed9568f4695ccd2e3a3