Infrared spectra of microbial cells are highly specific, fingerprint-like signatures which can be used to differentiate microbial species and strains from each other. In this study, the potential applicability of Fourier transform infrared (FTIR) spectroscopy for the classification of yeast strains in terms of their biological taxonomy, their use in the production of wine, beer, and bread, and their sensitivity to killer yeast strains was investigated. Sample preparation, spectral data preprocessing methods and spectral classification techniques were also investigated. All yeast strains were grown on a single growth medium. The FTIR spectra were baseline corrected and the second derivative spectra were computed and employed in spectral analysis. The classification accuracy was improved when the principal component spectra (calculated from the second derivative spectra) were employed rather than the second derivative spectra or raw spectra alone. Artificial neural network (ANN) with 10 units in the input layer and 12 units in the hidden layer produced a robust prediction model for the identification of yeasts. Cluster analysis was employed for the classification of yeast strains in terms of their use in the production of wine, beer, and bread and in terms of their sensitivity to killer yeast strains. The optimum region for the classification in the former case was found to be between 1300 and 800 cm-1 in the infrared spectrum whereas the optimum region for the classification of yeast strains in terms of their sensitivity was between 900 and 800 cm-1 . The results of this work demonstrated that FTIR spectroscopy could be successfully employed for the classification and identification of yeast strains with minimal sample preparation.

No description available.