RadBO · Cracking the Radiochemical Black Box with Bayesian Optimisation

Detecting disease at its earliest stages depends on being able to see what is happening inside the body at the molecular level. Positron Emission Tomography (PET) is a powerful imaging technique that makes this possible by using special radioactive molecules, called tracers, to reveal biological activity in real time. New technologies, like total-body PET, are revolutionizing PET imaging by allowing earlier detection and faster diagnoses, leading to more effective treatment for complex diseases like cancer, dementia, and heart disease. To unlock the full potential of total-body PET, we need new tracers that can target a wider range of diseases and biological processes. However, progress has been hindered not by a lack of promising tracer molecules, but by the difficulty of developing efficient radiochemical reactions to reliably produce them. This research will develop open-source machine learning tools to guide and accelerate these types of reactions. These tools will be validated using real radiochemical datasets, demonstrated on the synthesis of clinically relevant [¹⁸F]-radiotracers, and applied in the discovery of new radiochemical reactions. By doing so, this research will support a broad community of chemists, medical imaging specialists, and pharmaceutical scientists in harnessing the full potential of PET for studying and diagnosing disease.