In-depth analysis of the characteristics of volatile organic compounds in wines: a systematic study integrating intelligent sensory and metabolomics techniques with chemometrics and machine learning models

The volatile organic compounds (VOCs) in wines of 'Dornfelder' (DF), 'Petit Verdot' (PV), 'Pinot Noir' (PN), 'Sangiovese' (SV) and 'Malbec' (MB) were analyzed using an E-nose, HS-SPME-GC-MS and HS-GC-IMS. A total of 94 VOCs were identified by two techniques. Specifically, HS-SPME-GC-MS identified 70 compounds (alcohols' concentration accounting for 52.56%-68.75 %), and HS-GC-IMS identified 36 compounds (esters' concentration accounting for 35.58 %-42.05 %), with 12 compounds were identified by both methods. 15 key differential VOCs identified through chemometrics and machine learning analysis. Additionally, correlation analysis of E-nose sensor responses with key differential VOCs indicated that W2S, W2W, and W5S may be more suitable for predicting levels of 2-methylbutyl acetate, 3-methyl-butanoic acid, and isoamyl acetate, which can thus help to quickly identify PV wine. These results help to understand the flavor differences between different varieties of wines and provide a theoretical basis for wine flavor differentiation, quality control and product development.