Multivariate data analysis applied on Fourier Transform Infrared spectroscopy for the prediction of tannins levels during red wine fermentation
Red wine is a beverage with one of the highest polyphenol concentration, which are extracted during the maceration step of the winemaking process. Sensory perception such as astringency and bitterness are mainly related to tannin concentration and composition. However, quick analytical measurement of polyphenolic compounds can be a real challenge for monitoring their extraction during fermentation. Many methods were developed to analyzed polyphenols in wine, but they are time-consuming and require chemistry skills and equipment, not suitable for a rapid routine analysis. Thus, development of predictive models
using Fourier transform infrared (FTIR) spectroscopy coupled with chemometrics analysis appears to be a reliable and rapid method to determine polyphenolic content during wines fermentation.
For this purpose, this work sought to determine correlation between FTIR analysis and regular quantification methods for tannins, for different samples, covering three different vintages with two different grape varieties, from the beginning to the end of the extraction process. The search for diversity was highlighted during the selection of samples, to provide the best representation of the winemaking process. Total tannin concentration was analyzed by protein and polysaccharide precipitation. Flavanol composition was obtained by HPLC-UV after phloroglucinolysis reaction. FTIR spectra were registered between 925 and 5011 cm-1 using Winescan. Correlation between spectral analyzes and the various analytical information obtained were sought with partial least squares (PLS) multivariate regression analysis, for designing prediction models. The different models were tested with cross validation, and validation with an external set of samples to the calibration. For the external validation, the dataset was split into calibration and validation using Kennard-Stone algorithm.
The objective of this study was to demonstrate the interest of FTIR with PLS multivariate regression analysis to predict tannins concentration during winemaking. Correlations obtained show relevant results for the studied parameters, with models coefficients for cross validation higher than 0.8 for flavanol content (except for epigallocatechin) and higher than 0.9 for total tannins concentrations. The results with external validation are slightly lower for total tannins concentrations, with coefficient of prediction around 0.87, and show a more important decrease for flavanol content, with coefficient of prediction close to 0.7. If models for total tannins already show a high robustness and prediction, models for flavanol content must be improve with other samples. However, the results are encouraging and an
increase of the robustness could allow following flavanol content during winemaking. This work is the first step for the construction of predictive models to quantify different flavanol parameters in red wine fermentation by FTIR spectroscopy.
Issue: IVAS 2022
¹UR Œnologie EA 4577, Université de Bordeaux, ISVV
²USC 1366 INRAE, IPB, INRAE, ISVV
³FOSS Analytical A/
Contact the author
Tannins, Flavanol, Partial least squares regression, Fourier Transform InfraRed