Since culture-based methods are costly and time consuming, alternative methods are investigated for the quantification of probiotics in commercial products. In this work ATR- FTIR vibration spectroscopy was applied for the differentiation and quantification of live Lactobacillus (La 5) in mixed populations of live and killed La 5, in the absence and in the presence of enteric polymer Eudragit® L 100-55. Suspensions of live (La 5_L) and killed in acidic environment bacillus (La 5_K) were prepared and binary mixtures of different percentages were used to grow cell cultures for colony counting and spectral analysis. The increase in the number of colonies with added%La 5_L to the mixture was log-linear (r2 = 0.926). Differentiation of La 5_L from La 5_K was possible directly from the peak area at 1635 cm-1 (amides of proteins and peptides) and a linear relationship between%La 5_L and peak area in the range 0-95% was obtained. Application of partial least squares regression (PLSR) gave reasonable prediction of%La 5_L (RMSEp = 6.48) in binary mixtures of live and killed La 5 but poor prediction (RMSEp = 11.75) when polymer was added to the La 5 mixture. Application of artificial neural networks (ANNs) improved greatly the predictive ability for%La 5_L both in the absence and in the presence of polymer (RMSEp = 8.11 × 10-8 for La 5 only mixtures and RMSEp = 8.77 × 10-8 with added polymer) due to their ability to express in the calibration models more hidden spectral information than PLSR.
Keywords: ATR-FTIR; Artificial neural networks; Differentiation and quantification; Lactobacillus acidophilus La 5; Partial least square regression; Prediction models.
Copyright © 2018 Elsevier B.V. All rights reserved.