Batch‐to‐Batch Optimization of pH Shifts in CHO Bioreactor Operations

Culture pH is an important operating parameter in CHO cell cultures producing mAb since it directly affects growth and specific productivity. Mathematical models can be used to optimize the magnitude and timing of pH shifts provided that they are robust to model errors. To address model errors, we implement, for the first time experimentally, a robust batch-to-batch optimization procedure that iteratively adjusts the model by updating model parameters and by matching predicted to measured gradients of the objective function and constraints at each iteration. A dynamic metabolic flux model with pH-dependent kinetic constraints was used for optimization. The approach was applied experimentally using a system of micro-bioreactors, leading to an improvement of the titer up to 18% compared to the nominal experimental conditions. The necessity of gradient matching to compensate for model errors was clearly demonstrated by comparing experiments conducted with and without gradient matching. The mathematical model also allowed the identification of metabolic mechanisms that contribute to the observed improvements in productivity.