MRC Doctoral Training Partnership: Application of Machine Learning to Neuronal Phenotypic Models to Identify drug targets in Motor Neuron Disease

Artificial Intelligence has the potential to accelerate early drug discovery in a number of ways by making use of large data sets relating to heritable causes of diseases and gene expression profiles from in vitro and in vivo models and the human disease itself. Building computational in silico models of diseased cells and predicting responses to specific pertubations is one area that can help to identify druggable targets early in the discovery process. Training computational models to learn from wet biology experimentation in order to improve their prediction success in a variety of contexts is a major goal in the field of artificial intelligence (AI).

At the Sheffield Institute for Translational Neuroscience (SITraN http://sitran.org/) we have expertise in neurodegenerative models and drug discovery (Dr Richard Mead http://sitran.org/people/mead/) and AI/data science applied to disease process and drug discovery (Dr Dennis Wang http://sitran.org/people/wang/) as well as an ongoing collaboration with one of the UK’s leading Biotechs with expertise in G-protein coupled receptors (GPCRs), the most highly drugged class of proteins in the human genome. By bringing these three strands together, this PhD project provides a unique opportunity for the successful student to apply cutting edge informatics and artificial intelligence approaches to the problem of drug discovery, execute in vitro experiments to validate and improve their model and identify real targets which can enter the drug discovery process at an Industrial partner for which there will be placement opportunities.

The overall approach would involve finding single GPCRs targets or combinations of GPCR targets with downstream effects on cell fate that could have therapeutic potential, as well as finding unliganded and/or orphan (but predicted druggable) GPCR targets with similar potential. Techniques would rely heavily on machine learning based frameworks of pi-calculus (doi: 10.1186/1752-0509-3-118) and boolean networks ( doi: 10.1158/0008-5472.CAN-16-1578) to model biological signaling pathways in cell lines. The candidate must also be interested in in vitro experimentation in cellular models of human disease as they will be generating wet lab experimental data to validate and improve their models. They will also receive hands-on training and experience using machine learning software developed by our collaborators at Microsoft Research (http://biomodelanalyzer.org) to develop in silico cellular models integrating genetic and expression data. As such this is a unique opportunity to learn a number of cutting edge approaches to early drug discovery and apply them to a real world problem.