Hybrid modelling and data-driven parameterization of monoclonal antibody cultivation processes: Shifts in cell metabolic behavior

Process developments to achieve more efficient monoclonal antibody (mAb) production processes include cell line modifications to obtain high-performance alternatives. Cell lines with higher growth rates can exhibit higher sensitivities to variations in carbon sources in the cell culture leading to shifts in metabolic behavior. Current kinetic process models often rely on static model parameters like specific cell consumption and production coefficients. Such models can be inadequate for describing process performance in cases of shifts in cell metabolism. Further understanding of the complex biological phenomena is required to develop more robust process models. In this work, pilot-scale experimental results were obtained for a novel high-performance cell line exhibiting switches in metabolic behavior. State estimation-based methods were used to evaluate the model fit over the course of the run and to identify regions where parameter updates are required. Clustering of the underlying experimental conditions was performed to isolate the sources of variations and correlate them with the different regions of model fit. Alternative empirical formulations are accordingly proposed for the varying model parameters. This approach allows for the development of more interpretable hybrid models by using the data-driven insights to categorize and describe the underlying variations in operating parameters and their influence on cell behavior and process performance.