Health conditions are often complicated with many factors affecting how they may arise, develop, and be experienced by people. Traditionally the cost and complexity of gathering and analysing patient data at the necessary scale and detail needed is a barrier to understanding what is happening. This hinders the discovery of effective ways to intervene to improve outcomes for patients. Technical developments in medicine, biology, and computing in recent years have reduced many of these barriers so that it is now possible to measure many of the features of healthy people and those suffering ill health at scale. Examples would include the sequencing of people’s genomes, improvements in medical imaging techniques, or testing that can simultaneously measure levels of many molecules with high accuracy at low cost. The UK Biobank project hosts such data from 100s of thousands of people in the UK who have consented to have their data used to help research into ill health, but one major challenge is how to make best use of this valuable data. Over the next 3 years this project will build “patient similarity networks” that capture how similar any two patients are to each other based on the different types of data that are available in the UK Biobank. For example if two people have very similar medical histories, genetic data, or clinical assay results. We have developed novel methods to combine patient networks from multiple types of data and build machine learning models that capture relationships between patient features and ill health. This integrated network better captures similarity between patients and, because we are integrating very different types of data, captures aspects of ill health not accessible using one type of data alone. These networks will allow us to identify sub-groups of patients that are highly similar to each other and their associated features. These features could be used to develop bespoke management and treatment strategies and assist in identifying novel targets for the development of new therapeutics.
