Chemo-Sensory Markers for Red Wine Grades: A Correlation Study of Phenolic Profiles and Sensory Attributes

To reveal the characteristic physicochemical indicators of wines of different quality grades and explore their feasibility as auxiliary indicators for grading, 23 wines from the Manas subregion of Xinjiang were used as test materials. Sensory evaluation, colour difference analysis, and electronic tongue technology were employed, combined with nontargeted metabolomics and quantitative analysis, to analyze differences in phenolic compounds, colour parameters, and taste characteristics among wines of different grades. Finally, a quality evaluation model for Cabernet Sauvignon wine was constructed using partial least squares regression (PLSR). The results revealed significant differences in the L* values, a* values, and C*ab values among wines of different grades. Grade A wines presented lower L* values, higher a* values, and higher C*ab values, indicating lower brightness, deeper red tones, and higher saturation. Taste characteristic differences were primarily manifested in Grade A wines, which have higher acidity, astringency, bitterness, and richness but exhibit lower bitterness aftertaste and astringency aftertaste. The results of the quantitative analysis and correlation analysis indicate that the differences in sensory characteristics among different grades of wine stem from variations in their polyphenolic compound contents. The higher anthocyanin content in Grade A wine is associated with higher a* values: higher flavonoid content is closely related to higher astringency and bitterness values: and lower flavanol content is associated with lower bitterness aftertaste and astringency aftertaste values. The PLSR model results indicate that when sensory characteristic parameters and phenolic compound content are used as predictor variables (X) and grade is used as the response variable (Y), the PLSR model has a calibration set R2 = 0.97 and a validation set R2 = 0.92, the calibration set RMSE is 0.13, and the validation set RMSE is 0.25. The model demonstrates good fitting performance, establishing an objective method for evaluating wine quality that avoids evaluation errors caused by the subjective factors of winemakers and tasters. This study is the first to conduct a comprehensive evaluation of the sensory characteristic and chemical components of three grades of wine, providing data support and theoretical references for the improvement of wine quality evaluation systems.