Characterize the dynamic changes of volatile compounds during the roasting process of Wuyi rock tea (Shuixian) integrating GC-IMS and GC × GC-O-MS combined with machine learning.

Abstract

Understanding aroma compounds' changes during Shuixian roasting is vital for scientific guidance. This study used gas chromatography-ion mobility spectrometry (GC-IMS) and two-dimensional gas chromatography-olfactory-mass spectrometry (GC × GC-O-MS) to identify 100 and 183 compounds, respectively. The random forest algorithm combined with relative odor activity value (rOAV) was used to identify eight key differential compounds: 3-methylbutanal, (E)-2-octenal, 5-methylfurfural, 2-ethyl-5-methylpyrazine, 1-furfuryl pyrrole, 1-(1H-pyrrol-2-yl)-ethanone, 1-octen-3-ol, and (Z)-4-heptenal. The metabolic pathways of these compounds were analyzed, mainly involving the Maillard reaction and lipid oxidation. This study not only provides theoretical support for the targeted processing and quality control of Shuixian but also introduces a novel methodological framework by integrating advanced analytical techniques with machine learning, offering new insights into the dynamic changes of volatile compounds during tea roasting.