Volatile Organic Compound-Based Predictive Modeling of Smoke Taint in Wine.

Area of Science:

• Agricultural Science
• Food Science
• Analytical Chemistry

Background:

• Smoke taint significantly degrades wine quality, posing a critical challenge for the wine industry.
• Volatile organic compounds (VOCs) are key indicators of smoke taint in grapes and wine.

Purpose of the Study:

• To develop data-driven models for predicting wine smoke taint using VOC concentrations.
• To assess the feasibility of predicting smoke taint in grape and wine samples prior to fermentation.

Main Methods:

• Analysis of 20 VOCs in 48 grape and 56 wine samples.
• Application of univariate analysis and predictive modeling, including log conversion of VOC concentrations.
• Development of regression models for predicting smoke taint index and classification models for detecting tainted wines.

Main Results:

• Predictive models using both grape and wine VOCs achieved high accuracy (PCC = 0.82, R² = 0.68).
• Models using only grape VOCs also demonstrated strong predictive power (PCC = 0.76, R² = 0.56).
• Classification models effectively detected smoke-tainted wines with high precision (0.92) and F1 scores (0.82).

Conclusions:

• VOC concentrations in grapes and wine can accurately predict smoke taint levels.
• This predictive capability allows for early detection and management of smoke taint before fermentation.
• The study provides a data-driven framework and accessible code for smoke taint prediction in the wine industry.