Key aroma-active compounds and machine learning classification of chestnut baijiu revealed by molecular sensory, docking, and pathway analyses

Chestnut baijiu is a strong-flavoured spirit, distilled from steamed, Aspergillus-saccharified and Saccharomyces-fermented chestnut. To explore the composition and contribution of aroma, HS-SPME/LLE GC-O-MS and HS-GC-IMS were employed in this study to identify 235 VOCs in chestnut baijiu QD, QX, and JA from three representative chestnut-producing regions. 9 key aroma-active compounds, including heptanal, ethyl butanoate, (Z)-4-heptenal, (E)-3-hexenal, hexanoic acid ethyl ester, 3-furaldehyde, 2-pentanol, octanoic acid ethyl ester, and acetoin, were confirmed by odor activity values, docked into olfactory receptors with the lowest binding energy at octanoic acid ethyl ester, and their synthesis pathway were revealed. (E)-3-hexenal, originated from the isomerization of chestnut-specific (E)-2-hexenal and contributing a freshly cut grass aroma, was first identified in chestnut baijiu as a key biomarker among aroma-active compounds. 5 characteristic aroma-active compounds, 3-furaldehyde and hexanoic acid in QD, ethyl butanoate and hexanoic acid ethyl ester in QX, and acetoin in JA contributed pungent, fruity and sweet aromas, identified by the high flavour dilution (FD) factor values of AEDA. 8 differential aroma-active compounds, including (E)-3-hexenal, heptanal, propanal, 1-penten-3-ol, dodecanoic acid ethyl ester, octanoic acid, ethyl 3-methylbutanoate, heptanoic acid ethyl ester were screened through OPLS-DA and RF importance ranking. Another RF model achieved a classification accuracy of ≥0.95 on the feature set of differential compounds, enabling meaningful grading of the aroma of this alcoholic beverage. Machine learning based on aroma-active compounds provided higher chemical resolution and interpretability than e-nose sensor responses. Chestnut baijiu in sensory analysis had more pronounced malt aroma, with variations in daqu aroma among samples.