A hybrid modeling framework to optimize Chinese hamster ovary (CHO) cell cultures for monoclonal antibody (mAb) production has been developed, significantly reducing the number of modeling parameters while delivering results comparable to leading kinetic models. This innovative approach, created by researchers at Sartorius and the University of Waterloo, integrates dynamic flux balance analysis with piecewise partial least squares regression (dFBA-PLS), yielding a normalized mean squared error of 0.15 for most metabolites.
The implications of this framework are substantial for biomanufacturers, who often face challenges due to metabolic variability among CHO cells and the complexities of culture media. Traditional optimization methods can be labor-intensive, relying on a mix of empirical and mechanistic models. The dFBA-PLS model allows for direct modeling of media mixing ratios without requiring precise formulation details, enhancing the predictability of amino acid dynamics, glucose by-products, and biomass production.
In experiments involving 18 runs, the dFBA-PLS model demonstrated superior accuracy in predicting reaction rates compared to previous methodologies. This capability not only streamlines the optimization process but also facilitates in silico evaluations, potentially accelerating media development and reducing process variability. Furthermore, the scalability of this model suggests its applicability to other CHO DG44-based cell lines, marking a significant advancement in bioprocessing methodologies.
Use the database as your supply chain compass →
