Bioprocess analytics is needed to understand and control biotechnical processes. Different analytical systems, like optical sensors, spectroscopy (fluorescence, NIR, MIR, UV/Vis) and microscopic systems (In-situ-Microscopy) are used for online monitoring of process variables.
Bioprocess analytics is a main research area providing information to describe bioprocesses. Understanding of biochemical processes, as well as process control and product characterization are necessary for bioprocess analytics. Based on matrix complexity of a bioprocess, high process variability and mostly low concentration levels of the analytes are challenging to find an appropriate sensor for bioprocess monitoring.
Extensive infrastructure for online analysis of relevant chemical, physical and biological process variables is available at the TCI. Different spectroscopic methods (UV/Vis, MIR, NIR, and Fluorescence) enable non invasive measurements directly inside the bioprocess. New sensor concepts for measurement in small scale (96-well), shaking flasks (patches) or modern disposable bioreactors are available.
Analysis and interpretation of spectroscopic data by multivariate methods are the main focus of chemomtrics.
Chemometrics are used to get information out of online measured data. This Information can be used to monitor or to control the process. Factor analysis methods, like principal component analysis (PCA) or partial least square regression (PLS) are used to evaluate spectroscopic data completed with knowledge based methods. To monitor various process variables online, this technique is successfully applied for cultivation of different organisms in laboratory and industrial scale.
Next to the analysis of spectroscopic data Design of Experiments (DoE) is another chemometric tool. Herewith chemical experiments are designed by statistics to extract as much information out of as less as possible experiments.
Automation of bioprocesses
Aim of this research is to automate bioprocesses based on online estimation of the actual process state.
Data driven models are used to extract information from raw data resulting in online prediction of process variables or process trajectories. Knowledge based models are used to describe mechanistic processes to simulate the process or to forecast the process evolution. Both together enables an optimal feed rate to guarantee a high process yield.