Mass spectrometry similarity analysis and pattern recognition-driven metabolomics strategy for species and geographic chemotyping of Gardenia fruit.

Gardenia fruit possesses significant medicinal and edible value; however, its quality control remains challenging due to adulterant confusion and geographical diversity. Herein, an integrated mass spectrometry (MS1)/MS2 similarity filtering strategy including polygonal mass defect filtering, diagnostic ion and neutral loss analysis, and feature-based molecular networking was established to profile diverse chemicals of Gardenia fruit, preliminarily screening 3675/1942 flavonoids, 5642/2329 phenylpropanoids, 2502/1061 iridoids, 937/250 non-iridoid monoterpenes, 825/486 diterpenes, and 76/33 triterpene aglycones in ESI+/ESI- mode, and finally identifying 5 potentially new phenylpropanoid derivatives. Moreover, multivariate statistics and machine learning informed UPLC-Q-TOF-MS/MS-based untargeted metabolomics analysis were applied to compare metabolic differences and screen potential markers for their species authentication and geographic chemotyping. The key species marker set composed of d-mannitol, fumaric acid and asiatic acid generated a binary classifier with 100% accuracy that discriminated Gardenia fruit from its adulterants according to their characteristic abundance. And key geographic chemotyping markers were screened as genipin 1-gentiobioside combined with jasminoside B, a binary classifier built on their abundance also achieved 100% accuracy. This study offers robust technical and methodological support for the quality assessment, authentication and geographic chemotyping of Gardenia fruit.