Accurate and fast computational approaches to predicting chemical reactivity provide cost-effective alternatives to time-consuming experimental screening in synthetic route design. Transition state (TS) geometries are of great importance in determining chemical reaction mechanisms, rationalising reaction outcomes, and designing new transformations. However, current quantum mechanical methods for calculating TS geometries can be prohibitively expensive.
My research involves training machine learning (ML) models to predict TS geometries and offer a fast, accurate approach to providing mechanistic insight into the most widely used chemical reaction classes. These models will enable chemists to design new synthetic routes to cut down on wasted time and materials spent on failed experiments, optimise process conditions to result in more efficient and sustainable practices, and ultimately accelerate the development of new molecules. This application of ML represents a responsible use of AI aligned with the core values of the ART-AI CDT.
This project is part-funded by AstraZeneca.
Computational organic chemistry, quantum chemistry, chemical reactivity, catalysis, machine learning, deep learning, neural networks, software engineering.
MSci Chemistry at the University of Nottingham with an International Study Year at the University of Melbourne.
Graduate Scientist with AstraZeneca’s R&D Graduate Programme
Dr Matthew Grayson
