Comparing Kinetic versus Stoichiometric Priorities in Hybrid Models of CHO Metabolism

Understanding Chinese hamster ovary (CHO) cell metabolism through mathematical models is essential for optimizing culture media and biomanufacturing processes.

Current mechanistic models rely primarily on either flux balance analysis (FBA), estimating intracellular fluxes while assuming steady state, or kinetic modeling, capturing dynamic behavior but typically for a limited number of reactions. Dynamic FBA (dFBA) integrates both approaches in a hybrid framework, but challenges remain in integrating the two formats to describe bioprocesses. In this study, we first enhanced an existing dynamic CHO-metabolism model by incorporating ¹³ C-labeled data to refine kinetic expressions and stoichiometric constraints of amino acid pathways, including the asparagine-aspartate network and serine biosynthesis. We next evaluated the impact of prioritizing either stoichiometry, through the pseudo steady state assumption (PSSA), or the kinetic expressions of fluxes. Comparing error and predictive performance for both models for two industrially relevant fed-batch CHO culture conditions involving varying initial concentrations of nutrients and three feed streams, demonstrated that the kinetic-oriented model (KOM) yielded superior predictions for viable cell density (VCD), antibody production, and a range of amino acids and metabolites compared to the stoichiometric oriented model (SOM). Indeed, the KOM was able to predict production-to-consumption shifts of lactate and alanine, fluctuating levels of ammonia based on reversible kinetic expressions, and amino acids like asparagine and the serine-glycine pool. The KOM also provided better predictions for a third case including lactate-supplemented (LS) feed; however, slight parameter adjustments helped to improve model fidelity, likely due to the impact of high lactate on kinetic expressions of antibody (directly) and VCD (indirectly). In summary, our findings demonstrate that hybrid models emphasizing empirical kinetics over strict pseudo-steady-state constraints capture biologically realistic dynamics such as transient shifts for key metabolites like lactate, alanine, and ammonia, and also produce parameters useful across varying conditions, making them a practical and powerful tool for characterizing CHO cell culture performance in the future.