A multivariate regression approach based on sensory data and chemical compositions has been applied to study the correlation between the sensory and chemical properties of Pinot Blanc wines from South Tyrol. The sensory properties were identified by descriptive analysis and the chemical profile was obtained by HS-SPME-GC/MS and HPLC. The profiles of the most influencing (positively or negatively) chemical components have been presented for each sensory descriptor. Partial Least Square Regression (PLS) and Principal Component Regression (PCR) models have been tested and applied. Visual (clarity, yellow colour), gustatory (sweetness, sourness, saltiness, bitterness, astringency, and warmness) and olfactory (overall intensity, floral, apple, pear, tropical fruit, dried fruit, fresh vegetative, spicy, cleanness, and off-odours) descriptors have been correlated with the volatile and phenolic profiles, respectively. Each olfactory descriptor was correlated via a PCR model to the volatile compounds, whereas a comprehensive PLS2 regression model was built for the correlation between visual/gustatory descriptors and the phenolic fingerprint. "Apple" was the olfactory descriptor best modelled by PCR, with an adjusted R2 of 0.72, with only 20% of the validation samples falling out of the confidence interval (α = 95%). A PLS2 with 6 factors was chosen as the best model for gustatory and visual descriptors related to the phenolic compounds. Finally, the overall quality judgment could be explained by a combination of the calibrated sensory descriptors through a PLS model. This allowed the identification of sensory descriptors such as "olfactory intensity", "warmness", "apple", "saltiness", "astringency", "cleanness", "clarity" and "pear", which relevantly contributed to the overall quality of Pinot Blanc wines from South Tyrol, obtained with two different winemaking processes and aged in bottle for 18 months.
Keywords: multivariate regression; phenolic profile; pinot blanc; sensory analysis; volatile profile; winemaking.