Fluorescence analysis, in particular two-dimensional excitation-emission matrix (EEM) spectroscopy, is a sensitive in-line process monitoring tool in the biotechnological production. Before it can be widely adopted by the industry, fast effective algorithms for the analysis of massive and complex fluorescence data should be developed. Weak emission signals are prone to various interferences complicating the modeling, such as excitation light Rayleigh scatter. The scatter is usually considered as an unwanted background to be avoided or removed.
This work is focused on effective usage of the entire spectral information. It has been shown that scatter peaks coming from the excitation or an external light source can be used in the data analysis to improve the performance of a multivariate model. A new fluorescence Lighthouse Probe™ (LHP) has been applied in-line to monitor Saccharomyces cerevisiae fermentation in a lab reactor. Exploratory data analysis of two fed-batch cultivations has been performed using Partial Least Squares regression and Multivariate Curve Resolution. A simple algorithm for the resolution of process profiles and two-dimensional spectra of individual fluorophores from three-way fluorescence data in the presence of intensive scatter has been suggested and applied to the process diagnostics. Copyright © 2011 John Wiley & Sons, Ltd.