Development of generic raman models for a GS‐KO ^TM CHO platform process

The monitoring and control of bioprocesses is of the utmost importance in order to provide a consistent, safe, and high-quality product for consumers. Current monitoring and control schemes rely on infrequent and time consuming offline sampling methods, which inherently leads to some variability in the process which may impact the product quality profile. As part of Lonza's dedication to process analytical technology (PAT) initiatives this study evaluated the ability to generate generic calibration models, which are independent of the cell line, using Raman probes to monitor changes in glucose, lactate, glutamate, ammonium, viable cell concentration (VCC), total cell concentration (TCC) and product concentration. Calibration models were developed from cell culture using two different CHOK1SV GS-KO^TM cell lines producing different monoclonal antibodies (mAbs). Developed predictive models, measured changes in glucose, lactate, ammonium, VCC, and TCC with average prediction errors of 0.44, 0.23, 0.03 g L^-1 , 1.90 × 10⁶ cells mL^-1 , and 1.85 × 10⁶ cells mL^-1 , respectively over the course of cell culture with minimal cell line dependence. The development of these generic models allows the application of spectroscopic PAT techniques in clinical and commercial manufacturing environments, where processes are typically run once or twice in GMP manufacturing based on a common platform process. © 2018 American Institute of Chemical Engineers Biotechnol. Prog., 34:730-737, 2018.