Intensity prediction of typical aroma characters of cabernet sauvignon wine in Changli County (China)

https://doi.org/10.1016/j.lwt.2010.06.003Get rights and content

Abstract

In this paper, several linear multiple regression models of aroma descriptors were built from potential active odorants in Cabernet Sauvignon red wines in Changli County. The modified frequency (MF%) of ten aroma description terms in sample wines were evaluated by 30 panelists trained using the aroma standards of “Le Nez du Vin”. Aroma compounds of sample wines were detected by Solid Phase Microextraction-Gas Chromatography-Mass (SPME-GC-MS), and 65 aroma compounds were identified and quantified. Those aroma compounds with odor active values (OAV) > 0.5 were chosen to build regression models for the eight characteristic aroma terms. Finally, five models were developed for five typical sensory terms: Blackcurrant, Bilberry, Green pepper, Vanilla and Smoked. These models were related to 13 aroma compounds. These compounds included 3-ethoxy-1-propanol, phenethyl acetate, 4-terpinenol, 2-hexen-1-ol, di-tert-butyl-phenol, β-terpinenol, hexanoic acid, octanoic acid, ethyl myristate, ethyl 3-hydroxy butyrate, isobutyl alcohol and 4-methyl-5-butyl-2(3H)-furan. ANOVA statistical analysis indicated that all five models regressed at 95% significant levels. t detections of the models showed regression coefficients of 99% or 95% significant levels. Correlation coefficients between the measured and predicted Y ranged from 0.714 to 0.999.

Introduction

The flavor of the wine was found to be one of the most important attributes considered when buying wines. Over the past ten years, many aroma compounds in mono-variety wines and regional wines have been identified and quantified. Some researches have analyzed the flavor characters of compounds separated by GC column using Gas Chromatography–Olfactometry (GC–O) (Culleré et al., 2004, Lee and Noble, 2003). However, not all the aroma compounds detected by GC can produce flavor. Those with concentrations lower than their olfactometry threshold cannot be perceived by the human olfactory system. For this reason, in some of the literature, odor active value (OAV) was used to express the odor activity of one aroma compound in wine. OAV calculation measures the concentration and threshold of the odor in the same matrix. Only those odorants with an OAV > 1 can be perceived and contribute to the whole aroma of the wine (Allen et al., 1994, Kotseridis et al., 2000). For example, Culleré et al. analyzed the aroma compounds of six premiums red wines in Spain and found that 40 odorants had odor activity (Culleré, Escudero, Cacho, & Ferreira, 2004). Li et al. delineated the aroma compounds of Chardonnay white wine in Changli County (China), finding that thirteen impact odorants had OAVs > 1 (Li, Tao, Wang, & Zhang, 2008).

However, aroma compounds in wine are very complex. Many aroma compounds contribute to the whole aroma of a wine. Research on the aroma compounds’ contributions to wine should take into account not only the number of odorants and their relative importance but also the possible existence of synergetic interactions between those odors and with the matrix constituents (Guth, 1997, Li, 2006). We hypothesize that mathematical models could be built to describe the relationship between aroma descriptors and aroma compounds in wine. From these models, the intensities of some aroma characters could be predicted. To realize this proposal, three challenges had to be overcome. First, it was necessary to identify and quantify as many aroma compounds in wine as possible. Since the number of aroma compounds in wine is significant and the range of their concentrations is great, the method to detect the aromas must be both consistent and accurate. This is expensive and complicated. Second, data on the sensory analysis of wine must be obtained. A dependable sensory analysis was carried out by systematically trained panelists. Finally, the prediction result of aroma character from aroma compounds was a statistical estimation or inference value, so the wine sample number must be large enough for the prediction to be representative. For the reasons above, the prediction of aroma characters from aroma compounds is very seldom in the literature. In several previous studies, compounds producing some special flavor were discussed. Past studies have been limited to one compound or one group of compounds. Researches have quantified aroma compounds or described the flavor characters of them (Arrhenius et al., 1996, Aznar et al., 2003, Escudero and Ferreira, 2000).

In our work, sensory analysis and aroma compound detection were carried out in order to understand the aroma characters of Cabernet Sauvignon red wine in Changli County, one of the four districts of Wine Denomination of Origin in China. Our aim was to develop a method of mathematical prediction to describe the typical aroma characters of the sample wine. Linear multiple regression models were used to predict the intensities of aroma characters based on the contents of aroma compounds.

Section snippets

Wine samples

Wine samples investigated in this study were the same to that in the article wrote by Tao, Liu, and Li (2009). Eight different mono-varietal Cabernet Sauvignon wines from 1998–2005 were kindly donated by two wineries From Changli County. The winemaking process and some general indexes of wine samples were described by Tao et al. (2009).

Reagents

All reagents used were of analytical grade. Absolute ethanol, tartaric acid and sodium chloride were purchased from Xi’an chemical factory (Xi’an, China). Water

Sensory analysis

The sensory data were great. Some aroma terms with relatively low MF(%) values (<5.0) were omitted, leaving 32 observed aroma terms. The modified frequency (MF%) of these 32 aroma terms is shown in Table 1. Principle component analysis (PCA) was applied twice to obtain a more simplified view of the total aroma characters of the sample wines. The terms of blackcurrant, green pepper, smoked, redcurrant, cut hay, vanilla, bilberry and cinnamon most commonly described the typical aroma characters

Conclusion

Cabernet Sauvignon red wines in 8 years from Changli County were studied by sensory and chemical analyses. Five regression models of typical aroma terms of sample wines were built from potential active odorants. t Detections indicated that regression correlations and constants in the 5 models regressed at the significant level of 99% or 95%. Variance analysis showed that the 5 models regressed at the significant level at 99%. The correlation coefficients between the measured and predicted value

Acknowledgements

This project was supported by Elementary Science Research Fund of NWSUAF (QN2009061) and China National Science Fund (30571281). The authors are grateful to Huaxia Winemaking Company and Yueqiannian Winemaking Company (Changli County) for the supply of the samples used in this study.

References (30)

  • P. Chatonnet et al.

    The influence of Brettanomyces dekkera sp. yeast and lactic acid bacteria on the ethylphenol content of red wines

    American Journal of Enology & Viticulture

    (1995)
  • L. Culleré et al.

    Gas chromatography–olfactory and chemical qualitative study of the aroma of six premium quality Spanish aged red wines

    Journal of Agricultural and Food Chemistry

    (2004)
  • P.X. Etievant et al.

    Aroma components of pomaces and wine from the variety Muscat de Frontignan

    Journal of the Science of Food and Agriculture

    (1983)
  • A. Escudero et al.

    Clues about the role of methional as character impact odorant of some oxidized wines

    Journal of Agricultural and Food Chemistry

    (2000)
  • V. Ferreira et al.

    The aroma of Grenache red wine: hierarchy and nature of its main odorants

    Journal of the Science of Food and Agriculture

    (1998)
  • Cited by (127)

    View all citing articles on Scopus
    View full text