Mathematics for personalized medicine
Creating powerful mathematical methods for predicting the outcomes of diseases that pinpoint the right treatments and speed up drug trials.
Modern medicine can access vast amounts of patient-specific data, such as individual genomes, blood bio-markers and medical images. But it fails to use this information effectively. Analysing next-generation data with previous-generation tools causes avoidable patient harm, misses clues to new treatments and contributes to the high failure rate of clinical trials.
In this project we create more powerful mathematical methods for predicting disease outcomes. Most present-day mathematical and machine learning methods have a common Achilles heel. When the amount of information becomes too large, they confuse real patterns with noise and miss the wood for the trees—a breakdown known as overfitting. Our approach to bypassing this is founded on treating topological patterns in medical data as constraints on a random graph ensemble.
More accurate clinical prediction reduces the likelihood of harmful, ineffective treatments and saves lives. Each doomed trial that gets stopped early by better analysis of preliminary data saves over £100m for health services and the drug industry. And rescuing failing trials by identifying appropriate drug recipients broadens the range of available drugs.
Papers in this project
Bayesian networks describe the evolution of orthodontic features on patients receiving treatment versus no treatment for malocclusion.
Network analysis of diagnostic data identifies combinations of the key factors which cause Class III malocclusion and how they evolve over time.
Methods from tailored random graph theory reveal the relation between true biological networks and the often-biased samples taken from them.