Hybrid Modeling of CHO Cell Cultivation in Monoclonal Antibody Production with an Impurity Generation Module
Kozue Okamura, Sara Badr, S. Murakami, Hirokazu Sugiyama
Abstract
Representative cultivation models are needed for designing efficient monoclonal antibody (mAb) production processes. Simple Monod-type kinetic models could fail to capture changes at different phases and conditions. Current models rarely account for process-related impurities, which hinders optimizing integrated processes. A three-module hybrid approach is thus introduced here. Module 1 is a kinetic metabolism model until cell death based on Monod-type equations. Module 2 is a data-driven module for updating parameters based on system dynamics and changes in cultivation conditions. Module 3 proposes a kinetic model for the generation of host cell proteins and DNA. The model is applied to newly established high-performance CHO cell lines and is validated with experiments at different operating conditions. The adopted modular approach helps identify weak model links, which can be corrected via the data-driven module. The proposed model combines accuracy and simplicity for design and control applications of cultivation processes considering quality aspects.