Multi-strategy nonlinear model predictive control for Chinese hamster ovary continuous process

Abstract The biopharmaceutical industry has relied heavily on Chinese hamster ovary (CHO) cell lines to produce therapies using primarily fed-batch processes. Exploration of intensification strategies based on perfusion processes to increase volumetric productivity to reduce manufacturing footprint is an area of active investigation. Continuous perfusion processes provide significant increases in volumetric productivity. Operation of cell culture processes in a continuous mode presents unique automation challenges, to drive towards high yields while assuring cell health to enable extended run durations. Typical operating strategies for continuous bioprocesses include the use of a bleed stream to remove cells to maintain a target viable cell density known to provide stable cell health while maintaining a perfusion rate sufficient to supply required substrates. Here we propose a multi-strategy nonlinear model predictive control (NMPC) to allow switching between objectives ranging from stable operation to economic optimization. Economic optimization is primarily achieved through allowing the controller to operate at elevated cell densities and is enabled through use of the cell culture model that tracks accumulation of biomaterials and toxins and their influence on cell growth and death dynamics, allowing accurate prediction of maximum stable viable cell density. It is shown that, compared to a standard reference recipe, economic optimization results in over 68% increase in production rate while maintaining a viability above 95%. Validation studies are included demonstrating robust performance of the controller under plant model mismatch scenarios.