Run-to-run optimization of CHO cell culture media using high-throughput microscale bioreactor system and a hybrid modeling approach

Cell culture media composition plays a significant role in the resulting growth rate and titer of CHO cells producing Monoclonal Antibodies (mAb). In this study, a method for rapid run-to-run optimization of culture media is presented based on a hybrid Partial Least Squares- Dynamic Flux Balance Analysis (PLS-dFBA) model for blending fractions of commercial media combined with a gradient correction operation. The hybrid model was used for predictions and for gradient matching to guide the search toward the true plant optimum. The optimization objective was mAb titer which is predicted by integrating a time-varying specific productivity q p ( t ) obtained from the Dynamic Flux Balance Analysis (dFBA) component of the hybrid model. To optimize the cell culture, three commercial media were blended together in different fractions, and the run-to-run procedure searches for the combination of fractions that results in the highest specific productivity. The proposed procedure led to significant improvements in the specific productivity (¿20%) as compared to the baseline corresponding to the best titer obtained with each individual commercial media. Furthermore, the crucial role of the gradient correction step to achieve the optimum was illustrated by comparing the results to an optimization procedure without gradient correction.