Comparative study on the differences of non-volatile and volatile compounds in ten edible flower teas: An integrated multi-omics, molecular-docking and machine-learning approach

Chinese flower tea is valued as a traditional functional beverage for its aroma and health properties. We applied an integrated approach combining intelligent sensory evaluation (E-nose/E-tongue), untargeted metabolomics, GC–MS, molecular docking, and machine learning to analyze ten edible flower teas: Prunus persica cv. Duplex, P. speciosa , P. conradinae , P. salicina , P. mume , P. × blireana 'Meiren', P. serrulata var. lannesiana , P. persica , Malus × micromalus , and P. cerasifera 'Atropurpurea'. Sensory evaluation identified PSF, MMF, and PSLF as superior (scores >84), correlating with sweet/umami profiles. E-nose and GC–MS revealed distinct volatile signatures, with PSLF's floral aroma linked to high sesquiterpene abundance. E-tongue detected pronounced bitterness in PPF and PMF, which molecular docking with T2R46 associated with compounds like gambiriin A1, despite low alkaloid content. Metabolomics identified 3,049 shared metabolites, and docking with umami receptors (T1R1/T1R3) confirmed O-acetyl-L-homoserine and flavonoid glycosides as key taste-active compounds. Bioactivity assays showed PSLF's strong antioxidant capacity (DPPH/ABTS IC 50 0.375/0.242 mg/mL) and PSF's potent α-glucosidase inhibition (IC 50 0.086 mg/mL). Random forest modeling highlighted quercetin derivatives and citrusin isomers as major bioactive contributors. This multi-omics profiling provides a scientific basis for understanding the traditional appeal of flower teas and supports their targeted use in functional beverages.