Paul Kirk’s research focuses on the development and application of methods to characterise and treat diseases with greater precision. Clinicians and scientists are now able to gather a huge amount of data about patients and their diseases, including data from high-throughput biomolecular profiling technologies. These high-resolution datasets offer significant opportunities to improve patient health, as we can use the information they contain to identify groups of “similar” patients, who have similar disease outcomes, and/or who are likely to respond to medicines in a similar way. Paul’s research seeks to identify these groups and determine how to allocate patients to them, so that clinicians can make more precise prognoses, and target treatments more effectively.
Given the vast quantities of patient data that are now becoming available, one important question is: how can we determine which of this information is clinically useful? Paul uses principled probabilistic models and techniques from computational Bayesian statistics to identify and integrate the useful information from different data sources, and thereby ensure that patients are stratified in a way that has clinical utility.
Paul’s methods have previously been used to identify molecular subtypes of brain cancer that are associated with different survival outcomes, and he is currently working to adapt and extend these methods to other cancers and poor-outcome diseases.