Interpretable Machine Learning Algorithms for Predictive Toxicology

EPSRC CASE Studentship. University of Sheffield and Syngenta Crop Protection Machine learning is increasingly used for decision making and molecular design in the pharmaceutical and crop protection sectors to reduce the extensive time, costs and attrition associated with the development of new chemical entities. The typical aim is to relate molecular structure to predicted properties such as biological activity and toxicity. While complex Machine Learning Algorithms such as Deep Learning and Random Forest have been shown to deliver good prediction performance, they yield so-called “black-box” models which are challenging to interpret. Recently various approaches have been reported for interpreting the predictions of “black-box” models with varying degrees of success. An alternative approach has been the development of “interpretable-by-design” methods which, although they may have reduced overall performance, are by their nature easier to interpret, hence provide greater confidence in a regulatory context and better support chemists optimising molecular structures. Interpretability is particularly important in crop protection when trying to design out potential issues with (eco-)toxicology, to avoid late-stage attrition. The aims of this project are: the development of Machine Learning prediction methods that are both accurate and interpretable; and the extension of these methods to interspecies predictions to allow chemists to comprehend the reasons for species sensitivity. This studentship opportunity is only open to home (i.e. UK) candidates. Entry requirements are a minimum 2.1 undergraduate honours degree and/or MSc degree in a relevant science or engineering subject. It will be an advantage if candidates have an interest in Machine Learning or artificial intelligence and computer programming skills. The student will spend a placement period working at Syngenta Crop Protection. For more details on the entry requirements and research at the University of Sheffield, and how to apply, visit our department’s webpages at www.sheffield.ac.uk/is/phd Funding Fully funded 4 year studentship covering Home tuition fees, and an enhanced stipend for 4 years. The stipend pays the basic UKRI rate plus (currently £18,662 per annum for 23/24) plus an additional £4,000 per annum. There is also a generous research training and support grant to fund costs associated with the project.