Exploring Bayesian methods in chromatographic development: Increasing the capacity of the mRNA affinity ligand

• AI-based methods can be applied to chromatography optimisation reducing labour and cost. • Bayesian optimisation was used to increase resin’s dynamic binding capacity for the first time. • Bayesian optimisation increased DBC, up to 7.5 fold, in oligo-dT ligands. • Critical process parameters can be defined by the use of the Bayesian optimisation. Enabling the full potential of a new therapeutic modality requires the development of a flexible and cost-effective manufacturing platform. A critical bottleneck in this process is the development of robust purification platform, usually reliant on sequential chromatography steps. Development of chromatographic steps is a laborious and costly task, as it is dependent of multiple iterations. Evaluating multiple parameters in a traditional methodology can lead to an exponential number of experiments. To overcome this, we propose the use of Bayesian optimisation to efficiently navigate the solution space. Using the development of mRNA affinity chromatography as a model, Bayesian optimization was used to enhance the dynamic binding capacity. This approach led to a 7.5-fold increase in capacity (1.8 mgRNA mL -1 ) relatively to the benchmark run in only 13 iterations. Additionally, model interpretability techniques were used to correlate predictions with the experimental results, while gaining process knowledge. Bayesian optimisation is a powerful and efficient tool for chromatography development, and in combination with model interpretability techniques, can have a real impact on process development using a QbD framework, and with potentially be used for automation and broader application in bioprocessing.