Decoding the Flavor Structure of Jiang-Flavor Low-Alcohol Base Baijiu: A Machine Learning-Driven Approach to Reveal...

Area of Science:

• Food Chemistry
• Sensory Science
• Data Science

Background:

• Low-alcohol base baijiu, a byproduct of Jiang-flavor baijiu, has complex flavor profiles that are not fully understood.
• This limits its potential for high-value applications.

Purpose of the Study:

• To characterize the flavor structure of low-alcohol base baijiu using flavoromics and machine learning.
• To identify key factors driving its flavor profile and sensory attributes.

Main Methods:

• Flavoromics combined with interpretable machine learning (Random Forest, XGBoost, SHAP analysis).
• Analysis of flavor compounds, ethanol content, acidity, and pH during distillation.
• Sensory evaluation of baijiu samples across different alcohol concentration gradients.

Main Results:

• Flavor profiles systematically evolved with decreasing ethanol content, showing decreased esters/alcohols and increased acids.
• 3-methyl-1-butanol, furfuryl ethyl ether, and tetramethyl pyrazine were key markers for desirable flavors.
• XGBoost model accurately predicted flavor stages (95.5% accuracy), with SHAP analysis pinpointing esters and acids as critical regulators.

Conclusions:

• Established a quantitative and interpretable framework for understanding low-alcohol baijiu flavor.
• Identified key compounds and their impact on sensory attributes, enabling precision blending and quality grading.
• Provides a basis for the valorization of low-alcohol base baijiu.