Funded PhD scholarship:
'Understanding artificial intelligence models for toxicity prediction'

Scholarship details:

  • Qualification type: Funded PhD project
  • Funding amount: £15,009 stipend per annum plus tuition fees
  • Hours: Full-time
  • Closing date: 30 April 2019


  • IELTS 6.5 with 6 in each component OR a recent degree from a majority-English-speaking country
  • MSc with merit
  • UK or EU nationality
  • Suitable to computer science students with an interest in chemistry and toxicology, and chemistry students with an interest in computer science and machine learning.

Project description

The Sheffield Chemoinformatics Research Group is pleased to offer a fully-funded three-year PhD studentship in collaboration with Lhasa Limited. The student will be based at the University of Sheffield’s Information School (ranked best in UK and Europe in its field by the QS University ranking 2019) with an expected placement at Lhasa Limited’s base in Leeds, a world leader in knowledge and data sharing to improve the drug development process.

Artificial intelligence (AI) models, such as those obtained from deep neural networks (DNN), are seeing increasing usage in the chemoinformatics field. However, so far, applications of DNNs have focused on improving the accuracy of models, and comparing them to state-of-the-art methods. There has been comparatively limited research to understand the fundamental characteristics of the models trained. Thus, DNNs are considered ‘black box’ models because it is difficult to understand how they produce their predictions.

Computational prediction methods are of particular relevance for toxicity prediction. A key factor for the acceptance of predictive models for regulatory approval is the requirement to provide a mechanistic interpretation of the model. This requirement therefore limits the use of AI techniques for toxicity prediction.

In this PhD, the successful applicant will focus on an important aspect of DNN models that has seen little research so far: enhancing the interpretability of the trained models. They will apply different strategies with the aim of understanding which chemical patterns the models learn when they are making toxicity predictions. The results of the project could provide a deeper understanding of the causes of toxicity. The methods developed will be readily transferable to other types of chemical endpoints of interest such as biological activities.

UK/EU applicants will be eligible for a full award paying fees and maintenance at standard Research Council rates. The stipend RCUK rates for 2019/20 studentships are fees £4,327, stipend £15,009 per annum.

Application process
Prospective applicants should send their CV and two reference letters to or Pre-selected candidates will be invited to an interview.