Abstract

The study of human health and disease has been revolutionised by the new Omic technologies which are generating a high dimensional picture of the complex and confounding processes that occur during the initiation and progression of disease. Faced with the ability to generate bewilderingly large datasets, drug discovery researchers are now faced with a unique interpretation challenge. Machine Learning (ML) and Artificial Intelligence (AI) methods are garnering considerable interest as novel tools to support drug discovery and repositioning, enabling a synthesis of biologically meaningful signals from a background of data noise. Here the uses of machine learning in drug discovery will be described in several areas. Firstly a tensor-flow framework is described for the identification of targets and repositioning opportunities from genome wide association study data. Moving to human population genetics, we are using machine learning to predict benign human knockouts from genome sequencing projects, this information is highlighting safer targets with fewer side effects. Finally using gene expression data we are investigating predictors of response to biologic therapies. These diverse data sets are unified by a common machine learning analysis framework, that promises new insights into disease and drug discovery.