Leveraging statistical “Design of Experiments” (DoE) with high-throughput experimentation workflows to accelerate the development of novel radiopharmaceuticals

Introduction: Ready and rapid access to novel radiopharmaceuticals is crucial for the advancement of precision theranostics. Unfortunately, the traditional "low-throughput" radiosynthesis development cycle is often slow and costly. To address this, we aimed to develop a workflow to leverage the statistical "design of experiments" (DoE) approach to process optimization with high-throughput experimentation (HTE) protocols developed in our lab to explore radiochemical reaction space efficiently and to optimize difficult radiosyntheses systematically and rapidly. ^1,2

Methods: An HTE protocol using commercial equipment was developed to perform, analyze, and optimize multiple miniaturized (75-100 µl) Cu-mediated radiofluorination (CMRF) reactions simultaneously in glass micro vials in 96-well aluminum heating blocks.¹ The reactions were analyzed by separating ¹⁸F-labeled organic fractions from unreacted ¹⁸F^- using 96-well SPE cartridges. Activity concentrations of the organic and inorganic fractions were quantified via PET scanner and gamma counter,² and used to calculate the radiochemical conversion for each well. This HTE approach was refined for DoE optimization by incorporating an azeotropic drying-free ¹⁸F process.^{3 18}F^- was trapped on a QMA cartridge preconditioned KOTf and eluted with TBAF in methanol. The resulting [¹⁸F]TBAF solution could be easily distributed in 30-50 µl aliquots into 24- or 96-well plates and evaporated to dryness (100 °C, 3 min). Any reaction mixture, regardless of solvent, could then be added to the [¹⁸F]TBAF residue. JMP statistics software was used to plan miniaturized DoE experiments to study the CMRF of an [¹⁸F]crizotinib boronate precursor only available in very limited quantities.⁴ 24-well plate DoE studies were done and analyzed by rTLC. Each reaction was set up at one-tenth of a typical production scale experiment and performed in parallel (120 °C, 30 min) with stirring before reaction analysis.

Results: The HTE analysis protocol was used to study CMRF in 96-well plates. %RCC data (rTLC) correlated strongly to the PET (R²=0.972) and gamma counter (R²=0.974) data, which also exhibited low error (RMSE=5.8% & 5.9%,respectively) validating the protocol for high throughput reaction analysis. For 24-well plate studies, rTLC was a more efficient analysis method. Low-res DoE pilot studies conducted using a model precursor ((2,4-dichloro-3-methylphenyl)BPin) to screen categorical variables (solvents, ligand additives) and Cu(OTf)₂ loading identified imidazo[1,2-b]pyridazine (IMPY) and DMI as optimal ligand/solvent for the model precursor. The results informed design of a high-res 24-run, 4-factor D-optimal optimization study to model the effects of Cu(OTf)₂ (1-5 µmol), crizotinib precursor (0.25-2 µmol), and IMPY (1-40 µmol) loading and % n-BuOH co-solvent (0-25%). Full-scale validation using the optimal conditions predicted by the DoE response surface model afforded [¹⁸F]crizotinib in 57% RCC (n=1; predicted 55%). An alternative set of suboptimal conditions requiring less than half the amount of rare precursor also afforded the product in acceptable 40% RCC (n=1, predicted 36%). The optimized reaction is being automated on a GE TRACERlab.

Conclusions: A parallel high-throughput radiochemistry platform using common commercial equipment was successfully adapted for use with DoE response surface optimization studies. The method was applied to find effective radiolabeling conditions for [¹⁸F]crizotinib. The miniaturized HTE protocol allowed the entire 24-run DoE response surface optimization study to be carried out in a single 3-hour experimental session using just 27.8 µmol of the valuable and limited precursor. This study is a powerful demonstration of how new methods like DoE can be combined with HTE to drastically accelerate radiotracer development.

Refs: 1. Webb et al, Nuc Med Bio 2023; 2. Bowden et al, Sci Rep 2019; 3. Bowden et al, Org Bio Chem 2021; 4. Sardana et al, Pharmaceuticals 2022.

• Download figure
• Open in new tab
• Download powerpoint