Cabernet Sauvignon and Chardonnay grape sensory attributes and chemical compositions were mined to predict relationships with wine sensory characteristics. In parallel, winemaker panels defined sensory characteristics associated with quality Cabernet Sauvignon or Chardonnay wines. Relating the sensory characteristics of Cabernet Sauvignon or Chardonnay grapes to wines was challenging, as was modelling the entire wine sensory space using grape measures. However, modelling individual wine characteristics successfully linked blocks of grape measures to wine attributes. Knowledge generated from this project will form the basis for future development of measures of grape flavour potential and strategies to produce fit for purpose fruit.
Considerable research into the chemical basis for sensory attributes in wine has been undertaken but there has been less focus on understanding the links of grape composition to wine chemistry and wine sensory properties. In limited cases, a wine sensory attribute can be assigned directly to a specific grape metabolite; for example, pepper due to rotundone, green capsicum resulting from 3-isobutyl-2-methoxypyrazine, and floral characters from monoterpenes. Beyond that, sugars, amino acids, lipids, micronutrients, and other grape constituents will contribute to the suite of compounds produced during winemaking that are important to wine sensory properties. An increasing body of evidence emphasises the importance of grape composition on the potential of a wine to have certain sensory characteristics. Nonetheless, the previously identified gap in the literature still remains, and there is a lack of knowledge that explains how parcels of grapes from the same variety, and possibly same vineyard, can result in very different wine sensory outcomes. In addition, there is little information tracing vineyard management practices and effects of the environment on production of grape metabolites that subsequently influence wine chemistry and sensory. With regards to “quality” measures, some success has been achieved with red grape colour and this project aimed to develop other measures that predict wine sensory outcomes.
The project methodology consisted of three main parts. 1) In each of the first three vintages of the project (2013-15), 25 Cabernet Sauvignon grape parcels were obtained from our industry partners from both warm and cool growing regions and vinified using an identical small-scale fermentation protocol. A subsample of each grape parcel was analysed using multiple methods to quantify various classes of compounds including: amino acids, volatile compounds, bound volatile compounds, anthocyanins, flavonols, tannins, fatty acids and total phenolics. Other measures such as CIELab colour, normal harvest parameters and activities from lipoxygenase pathway enzymes were also conducted, as well as berry sensory analysis (BSA). The corresponding wines were analysed by a trained descriptive analysis (DA) sensory panel and rated for quality by a winemaker panel. The multiple datasets were then analysed for predictive relationships. 2) A similar experimental setup was applied to Chardonnay for two vintages (2015-16). Again, 25 grape parcels were obtained each year from our industry partners in vineyards spread across South Australia. These were vinified using a controlled protocol, and the wines profiled by DA and assessed for quality by a winemaker panel. BSA was conducted on the grape parcels and measures deemed to be relevant to Chardonnay fruit composition were applied to the fruit. Predictive relationships among the data were explored in a similar way to Cabernet Sauvignon. 3) Biochemical methods were used to study the pathways of two classes of secondary metabolites during Cabernet Sauvignon berry development, to explore whether genetic and biochemical markers could indicate changes in berry metabolism.
The study of Cabernet Sauvignon over three vintages showed that the major driver of differences in the sensory attributes of the wines was the region of origin. In general, the wines from the Riverland had lower sensory scores for dark fruit flavour and aroma, body, overall flavour and aroma and astringency, compared to the wines from other regions. These same wine sensory characteristics were identified by the winemaker sensory panel as indicators of higher quality, whereas lower quality wines were described as green, simple and poor in colour. Interestingly, some Riverland Cabernet Sauvignon wines that were graded higher than others, possessed “balance”, which is a holistic sensory percept that the detailed sensory profiling did not capture. To achieve a more complete understanding of wine quality, the drivers for concepts such as balance (and complexity), and not merely a list of specific wine sensory attributes, clearly require further investigations.
Because the study involved such a large number of data sets collected during an extensive metabolomics analysis of the grapes (12 data blocks for Cabernet Sauvignon, 9 for Chardonnay), a novel data analysis method called sequential and orthogonalised partial least squares (SO-PLS) was used to select data blocks that were predictive of wine sensory attributes. It also highlighted the data blocks that were least frequently used to model sensory perception and can likely be removed from future studies. The current study confirmed previous findings that some grape volatile measures are important for modelling sensory attributes in both Cabernet Sauvignon and Chardonnay wines. However, many other data blocks, arising from the quantification of groups of grape metabolites beyond grape volatiles, were better at modelling a range of wine sensory characteristics, and novel correlations between particular sensory attributes and blocks of grape metabolites were demonstrated. The grape target compounds may not necessarily be precursors to wine aroma volatiles, but may act as markers that indicate altered berry metabolism and composition. These grape biochemical markers of wine sensory outcomes would be useful in streaming or grading fruit once suitable protocols for their measurement can be developed and verified.
Winemaker quality assessment proved successful with the Cabernet Sauvignon wines but did not significantly discriminate the Chardonnay wines. When vinified with a simple, identical protocol, sensory differences across Chardonnay wines were detectable albeit subtle, and these results suggest that quality drivers of commercial Chardonnay wines are most likely not derived solely from grapes. Sensory differences were not consistent across the vintages and often did not relate to regionality. All nine blocks of grape measures were used in the models developed to predict wine sensory attributes so there appear to be many potential indicators of Chardonnay wine flavour. Our results suggest that vinification factors likely contribute more to the variation in sensory characteristics of commercial Chardonnay wines than the grapes, and perhaps grape measures of quality would be more suited to red varieties or other white varieties with distinct varietal characteristics.
Principal component analysis (PCA) of the data blocks enabled observation of the year to year variation in grape measures, yielding information about their stability across regions and vintages. This enables the identification of aspects of grape composition that can potentially be manipulated in the vineyard and those that may be more prone to variability due to unknown or uncontrollable environmental factors. Amino acid composition of the Cabernet Sauvignon grape samples was similar in samples taken from the same vineyard across the vintages, and was not primarily driven by region. This suggests that something intrinsic to the vineyard influences grape amino acid composition but exploration is required to determine whether this can be managed. Other variables, such as anthocyanin or tannin composition, appeared to be driven mainly by region, suggesting that broad climatic or regional management differences may be important determinants. Bound volatile compound and fatty acid compositions were somewhat related to region but they also varied from year to year. Their concentrations may be altered by environmental variables and could be managed if conditions in the bunch zone could be altered to mimic environmental changes.
Studies of the expression of genes from the lipoxygenase pathway and those responsible for the breakdown of carotenoids showed that the enzyme activity in the fruit was often the result of multiple genes and that maximal gene expression was often separated temporally from the peak in enzyme activity. This would make it difficult to use gene expression assays to predict pathway flux at harvest. Nevertheless, understanding where and when these pathways are active in the fruit is important for the development of strategies to manage the production of important aroma precursor compounds in the vineyard. This has important implications given the outcome that carotenoid content in Cabernet Sauvignon may predict the concentration of β-damascenone, a compound implicated in red wine quality.
The range of techniques and different intellectual approaches from analytical and separations chemists, plant physiologists, biochemists, oenologists and sensory scientists gave this project a genuinely multidisciplinary approach. The research team’s insights into grape chemistry have produced some novel and exciting results and have established a foundation of basic research that will eventually lead to changes in the way grapes are assessed and ultimately grown.
This study has brought together the complementary skills of three research teams at the University of Adelaide and CSIRO and we have identified a number of grape data blocks that are associated with specific Cabernet Sauvignon and Chardonnay wine sensory attributes. This is a significant step towards developing objective means of predicting wine sensory properties from grape measures.