Hybrid Model Captures Lactate Run-Away and Back-Metabolism

A new hybrid mechanistic-data-driven model of lactate consumption and production for mammalian cell cultures captures both lactate back-metabolism and lactate run-away. It also provides accurate real-time predictions of viable cell density, monoclonal antibody, glucose, and lactate concentrations during bioprocessing, according to researchers from University College London and AstraZeneca.

For biopharmaceutical manufacturers, this model, outlined in a recent paper, provides another tool to accurately predict batch-to-batch variability and failing runs early in the process, according to a team led by principal investigator Li Qin, PhD, postdoctoral research fellow, and senior researcher Stephen Goldrick, PhD, associate professor, digital bioprocess engineering, both at University College London.

By fine-tuning the initial model, the team improved upon the original 88% accuracy rate, reducing the root mean squared error by about 50% for glucose and by about 30% for lactate.

“This model addresses a key industrial challenge,” Goldrick tells GEN. “In fed-batch CHO cultures, cells normally switch from lactate production to lactate back-metabolism. If they do not, the culture goes into lactate run-away, which can cause batch failure or a significantly lower titer. This switch is poorly understood and rarely controlled.”

Therefore, a lactate switch detection algorithm is one of the key features of this model. As Goldrick explains, “Our model uses pH as a soft sensor to detect early, at around day 5, whether the cells will undergo lactate back-metabolism. This information could be built into the control strategy, so the process can be adjusted to avoid lactate run-away.”

Stratifying parameters

The novelty of this approach is three-fold, they say, encompassing “new kinetic equations to accurately describe Chinese hamster ovary cell growth and death dynamics…new equations for lactate production and consumption…to better capture metabolic transitions under varying process conditions…and adaptive kinetic updating integrated with a lactate-switch detection algorithm to incorporate online and offline measurements into the model dynamically.”

The model was trained with online data such as pH values, glucose flow rates, and other nutrient feeds combined with historical data such as concentrations of glucose, lactate, and mAb concentrations as well as cell densities accumulated from random runs. The team trained the model on five runs that switched from producing lactate to consuming it (called back-metabolism), three to evaluate predictive performance, one that exhibited lactate run-away (in which the cell culture continued producing lactate), and one external lactate run-away scenario.

That combination helped the researchers determine which parameters could be predicted accurately with online data and which highly variable parameters needed to be fine-tuned using a combination of real-time and historic data. “This targeted strategy significantly reduces model update time compared to full recalibration, while maintaining predictive accuracy,” they note.

Ten is a relatively small number of training runs, the scientists admit, but the values predicted by the model in those runs closely matched those of actual measured values, suggesting this model can make a real contribution to the industry. Further work will extend the measured production values to include amino acid consumption, ammonia accumulation, osmolality dynamics, and product quality attributes.

The post Hybrid Model Captures Lactate Run-Away and Back-Metabolism appeared first on GEN - Genetic Engineering and Biotechnology News.