Evaluation of spectrofluorometry as a tool for estimation in fed-batch fermentations

Area of Science:

• Biotechnology
• Biochemical Engineering
• Process Analytical Technology (PAT)

Background:

• Fed-batch fermentation requires accurate monitoring of biomass and glucose concentrations for optimal control.
• Traditional soft-sensors often rely on parameters like oxygen uptake rate (OUR) and carbon dioxide evolution rate (CER).
• Native culture fluorescence offers a potentially novel, non-invasive data source for process monitoring.

Purpose of the Study:

• To investigate the utility of native culture fluorescence as an additional data source for predicting biomass and glucose concentrations.
• To compare the performance of fluorescence-based soft-sensors with existing models.
• To evaluate the application of Partial Least Squares (PLS) regression and Feed Forward Neural Network (FFNN) coupled with Principle Component Analysis (PCA) for fluorescence data analysis.

Main Methods:

• Utilized native culture fluorescence measurements from fed-batch fermentations of Alcaligenes eutrophus.
• Employed Partial Least Squares (PLS) regression and FFNN-PCA models for kinetic analysis.
• Trained models using data from three fermentations and tested on a fourth, comparing with a conventional soft-sensor model.

Main Results:

• Native fluorescence-based models demonstrated superior performance in predicting both biomass and glucose concentrations.
• PLS and FFNN-PCA models utilizing fluorescence and its rate of change accurately predicted biomass in the testing set.
• FFNN-PCA models accurately predicted glucose concentration using the rate of change of fluorescence.
• A first-order PLS model using the rate of change of fluorescence qualitatively indicated substrate exhaustion.

Conclusions:

• Native culture fluorescence provides valuable predictive information that is not captured by conventional measurements.
• Fluorescence-based soft-sensors, particularly FFNN-PCA, show significant promise for enhancing fed-batch fermentation monitoring and control.
• This technique offers a cost-effective and non-invasive approach to improve bioprocess understanding and optimization.