Positron emission tomography (PET) serves as an invaluable asset in medical research and drug development, where physiological activities can be imaged non-invasively. Amongst the radioisotopes used in PET radiotracers, carbon-11 is highly valued for advantages such as its short half-life (20.4 min) being ideal for longitudinal in vivo studies.1 While its incorporation via methyl and carbonyl functionalities is well-established2, an ample knowledge gap remains for 3-membered rings (aziridines, cyclopropanes and epoxides), which are found in anticancer drugs including Olaparib. Nonetheless, a 2025 manuscript from Priest and co-workers have dawned this possibility through the Corey-Chaykovsky reaction (CCR), which returned promising radiochemical conversions.3 Similarly, this research seeks to optimize synthesis outcomes from a regioselective CCR reagent intended to minimize isotopic dilution otherwise expected with conventional sulfur ylides.
Where traditional OFAT/OVAT approaches leave researchers prone to falsely determined optimal synthesis protocols, Design of Experiment (DOE) avoids this issue through an initial exploration of parametric space aiming to maximize coverage while avoiding unnecessary sampling. A complete DOE campaign would have identified factors which significantly impacted reaction outcomes and along with prospective combinations. From there, the definition of these known optima regions can further be refined through a Bayesian Optimization (BO) algorithm, which uses deep-learning to determine new areas for sampling based on constructed acquisition functions. Ultimately, this work seeks to exemplify the benefits of using statistical algorithms by first understanding and fine-tuning optimal reaction condition sets over 3 model deuterated products, which can then be applied to the more expensive synthesis process of radiolabelling anticancer pharmaceuticals (e.g. Olaparib).