In silico dynamic optimisation studies for batch/fed-batch mammalian cell suspension cultures producing biopharmaceuticals

• Main Author: Lam, Ming-Chi
• Other Authors: Mantalaris, Athanasios ; Pistikopoulos, Efstratios
• Published: Imperial College London 2008
• Subjects: 615.19
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.505300

Mammalian cell cultures are valuable for synthesis of therapeutic proteins and antibodies. They are commonly cultivated in bioindustry in form of large-scale suspension fed-batch cultures. The structure and regulatory responses of mammalian cells are complex, making it challenging to model them for practical process optimisation. The adjustable degrees of freedom in the cell cultures can be continuous variables as well as binary-type variables. The binary-type variables may be irreversible in cases such as cell-cycle arrest. The main aim of this study was to develop a general model for mammalian cell cultures using extracellular variables and capturing major changes in cellular responses between batch and fed-batch cultures. The model development started with a simple model for a hybridoma cell culture using first-principle equations. The growth kinetics was only linked to glucose and glutamine and the cell population was divided into three cell-cycle phases to study the phenomenon of cell-cycle arrest. But there were certain deficiencies in predicting growth rates in the death phase in fed-batch cultures although it was successful to simultaneously optimise a combination of continuous and binary-irreversible degrees of freedom. Thus, the growth kinetics was further related to amino acids concentration and cellular responses to high versus low concentration of glutamine and glucose based on a Chinese hamster ovary cell-line where amino acids data were available. The model contained 192 parameters with 26 measured cell culture variables. Most of the sensitive parameters were able to be identified using the Sobol' method of Global Sensitivity Analysis. The model could capture the main trends of key variables and be used to search for the optimal working range of the controllable variables. But uncertainties in the sensitive model parameters caused non-negligible variations in the model-based optimisation results. It is recommended to couple such off-line optimisation with on-line measurements of a few major variables to tackle the real-time uncertain nature of the complex cell culture system.