A hierarchical model combining process and Raman spectroscopic data for real-time monitoring of cell culture process

Abstract

Process analytical technology (PAT) is expected to ensure process and product quality consistency through the timely measurement of raw and in-process materials and key process parameters in pharmaceutical production. With the application of multi-wavelength on-line spectroscopy such as NIR, UV andRaman, and the accumulation of on-line process data, multivariate data analysis (MVDA) has surpassed the traditional univariate or bivariate approaches asa more appropriate tool for bioprocess monitoring and characterization. In thisstudy, a multivariate hierarchical model combining on-line process and Ramanspectroscopic data was developed for real-time cell culture process monitoring.Non-invasive spectroscopy, such as NIR and Raman, can provide additional information to the process variable regarding the biochemical attributes of in-processmedia. The objective was to investigate the possible advantages of the hierarchical model using a combined data set compared to the model using process dataonly. A major advantage of the hierarchical approach was its ability to detectand interpret variations for in-process materials that cannot be detected by theprocess model. In addition, the hierarchical model enhanced the understandingof interactions between process and biochemical variables, as well as the impactof such interactions on process evolution. These advantages not only facilitate early fault detection, but also lead to a well-understood process and, ultimately, improved efficiency for process optimization and control.