The purpose of this research was to develop a rapid and simple method of identifying yeast and bacteria from wine using powerful new technologies, Raman spectroscopy coupled with multivariate classification tools. The research was conducted using pure cultures of 18 wine yeast and 18 wine lactic acid bacteria. Six strains of each of the following yeasts were analyzed: wine fermentation strains of Saccharomyces cerevisiae, and the wine spoilage yeasts, Zygosaccharomyces bailii and Brettanomyces bruxellensis. Spectra were acquired from 100 subcultures of each yeast strain and were analyzed using multivariate classification tools, Linear Discriminant Analysis (LDA) and Support Vector Machine-linear (SVML), a new machine learning technique. The accuracy of prediction of yeast identity to the strain level by LDA was 85%, although only approximately half of the spectra gathered per yeast were used due a problem with peaks generated by ambient light in the DeltaNu Raman spectrometer. The cross-validation accuracy of prediction of yeast identity to the strain level by SVML was 93%. Six strain of Oenococcus oeni, five wine species of Pediococcus, and six wine species of Lactobacillus were analyzed. Spectra were acquired from 24 subcultures of each bacterial strain and were analyzed using SVML. The cross-validation accuracy of prediction of bacterial identity to the strain level was 90%. The results of this research indicate that Raman spectroscopy coupled with a multivariate classification prediction may indeed be a simple way to quickly discover the identity of microorganisms in wine.
/wp-content/uploads/2017/09/AFV-Header-Logo.png 0 0 AVF /wp-content/uploads/2017/09/AFV-Header-Logo.png AVF2012-10-15 13:22:222019-01-22 18:49:38Detection of Brettanomyces, Zygosaccharomyces Bailii, Lactobacillus and Prediococcus in Wine Using Raman Spectroscopy