Cell culture processes are dynamic systems governed by a large number of influential and highly interacting process variables. Despite the stringent control of several process variables, considerable variation in the process outcome is often observed. Moreover, generalized deterministic models for dynamic process prediction based on the impact of the process variables on the product quantity and quality attributes do not exist. According to the Process Analytical Technology (PAT) initiative, multivariate data acquisition and analysis should be implemented to achieve advanced process knowledge and to gain effective control on risk in manufacturing. While several studies have shown that such multivariate statistical models provide promising possibilities for process monitoring and prediction based on the process history, they are very limited regarding their capability for the support of experimental design.
This work demonstrates how the flexibility of statistical models combined with the formalized bioengineering knowledge of simplified deterministic models can improve the efficiency in cell culture process development. Based on a rigorous case study, the results obtained with the hybrid approach will be compared to the ones from the purely statistical and deterministic models to visualize its promising capability of providing significant process information using the least number of experiments. Different kinds of hybrid cell culture models will be discussed and the possibility to incorporate such models into a sequential experimental design methodology to define the process design space will be presented.
See more of this Group/Topical: Pharmaceutical Discovery, Development and Manufacturing Forum