Kaiser hosts a series of complimentary webinar presentations which demonstrate Raman Spectroscopy-based solutions. Hear from real customers and their success stories using Kaiser Raman analyzers and instruments. Examples of Research Raman, Analytical Raman, Process Raman, In Situ Raman, Industrial Raman, Remote Raman, and Raman Imaging and Microscopy are available during these sessions. We invite you to register for an upcoming webinar of interest, or view previous archives. Kaiser Webinar Archives are available in the sidebar -- see what you've been missing!
Real Customers - Their Applications, Raman Solutions
Comparison of NIR and Raman Spectroscopy for Online Glucose Predictions Using a Shake Flask Model System
In this industry focused study, the capabilities of NIR and Raman spectroscopy were compared together by predicting the glucose concentration on-line, in the same CHO cell cultivations. By monitoring and controlling glucose concentration, the cell physiology is balanced that is positively correlates with produced protein quantity and quality. Glucose is the primary carbon source of the cells thus on-line monitoring and controlling of its concentration is crucial in bioprocesses.
A shake flask model system was used to generate on-line NIR and Raman spectra of cell cultivations to test the glucose predicting ability of the two methods and to develop a working model prior to bioreactor cultivations. A unique model building workflow was applied to test the effectiveness of different pre-processing and variable selection methods. Spectra of different shake flasks were divided to a calibration and a validation set and PLS regression models were developed using the calibration set. The glucose prediction ability of the models was tested on the validation part subsequently. Raman based PLS models could predict glucose in a shake flask with an RMSEP under 3 mM and to scale-up 10L and 100L bioreactors with an RMSEP of 4.0 mM.
In conclusion, the shake flask model system is capable of generating data for fast calibration that is also scalable, but only for Raman spectroscopy. Therefore, an accurate working model for monitoring the glucose concentration in a bioreactor during cultivation is available for the first bioreactor cultivation before any other bioreactor runs were conducted. This model could be developed to model predictive control with more bioreactor data.
AIR DATE: January 30, 2018
Name: Bence Kozma
Company: Budapest University of Technology and Economics