(207c) Eliminating Batch-to-Batch Variability in Monoclonal Antibody Production Using Closed-Loop Control | AIChE

(207c) Eliminating Batch-to-Batch Variability in Monoclonal Antibody Production Using Closed-Loop Control


Papathanasiou, M., Imperial College London
Kontoravdi, C., Imperial College London
Fed-batch bioreactors are the industry standard for the production of monoclonal antibodies (mAb) which are used for treatments of cancer. Most commercially available mAbs are produced by a mammalian cell-based system, namely Chinese Hamster Ovary (CHO) cells. Unlike bacterial cell cultures, mammalian cell-based systems allow for complex post-translational modification of the mAb to take place and lead to a more effective product. However, the complexity of these mechanisms and overall producing a product from a living organism leads to processes which exhibit increased batch-to-batch variability and inconsistent performance for example in productivity.

Current means of attaining a desired performance include open-loop control via optimisation of substrate feeding profiles (e.g., [1]). Unlike open-loop control, closed-loop control considers process feedback therefore compensation can be applied to account for process disturbances. Yet, there are very few examples of closed loop controllers for mammalian cell cultures. This is in part because of the highly non-linear nature of the process. In addition, process data is particularly expensive and laborious to generate. As such, model and control development remain challenging. Recent research in modelling and control of bacterial cell cultures has highlighted the advantages of closed-loop control (e.g., [2]). Similar advances are yet to be applied to mammalian cell-based systems.

In this work, an advanced control strategy is developed to track product titre by manipulating the feed of key metabolites to the cell culture. Using limited experimental data, a reduced-order model is developed for a controller describing only the manipulated, controlled, and constrained variables. From this, a model predictive controller is developed and the results from the simulation studies used to showcase the controller’s performance. Future implementation is expected to reduce variability between batches.

[1] - Kotidis, Pavlos, et al. "Model‐based optimization of antibody galactosylation in CHO cell culture." Biotechnology and bioengineering 116.7 (2019): 1612-1626.

[2] - Oh, Tae Hoon, et al. "Integration of Reinforcement Learning and Model Predictive Control to Optimize Semi‐batch Bioreactor." AIChE Journal (2022): e17658.