Quantitative analysis of yeast fermentation process using Raman spectroscopy: Comparison of CARS and VCPA for variable selection.

Abstract

Yeast is one of the most widely used microbial species in the field of microbiology, and it is crucial that rapid and accurate monitoring of its process. Therefore, this study presents a method using Raman spectroscopy for quantitative analysis of yeast fermentation process. First, a ProSP-Micro2000K Raman measuring system used to obtain the Raman spectra of eight batches of yeast samples during fermentation, and the spectra obtained were pretreated using Savitzky-Golay (SG) smoothing filter and standard normal variate (SNV). Then, two variable selection methods, which were competitive adaptive reweighted sampling (CARS) and variable combination population analysis (VCPA), were compared to search the preprocessed Raman spectroscopy characteristic wavenumber. Finally, support vector machine (SVM) was employed to construct a quantitative monitoring model of yeast fermentation process based on variables from the selected characteristic wavenumbers. The results revealed that the VCPA-SVM model showed the best prediction result with 14 selected characteristic wavelength variables. The coefficient of determination (RP2) of the optimal model was 0.979, while the root mean square error of prediction (RMSEP) was 0.108 in the validation set. The overall results demonstrate that the Raman spectroscopy integrated with chemometric approaches could be utilized as a rapid method to monitor the process of yeast cultivations.