Integration of UHPLC/QTOF-MS and GC-MS to decipher the subtle component difference between raw and bran-roasted Myristica fragrans (nutmeg).

To decipher subtle chemical variation of both non-volatile and volatile components during bran-roasting of nutmeg (Myristica fragrans), we innovatively established a "three-source comparison" approach (raw nutmeg, RN; bran-roasted nutmeg in lab, BN-L; bran-roasted nutmeg from market, BN-M) and developed chemometrics/machine learning (ML) models to identify processing-related markers. Untargeted high-resolution liquid chromatography-mass spectrometry-based metabolomic analysis of the RN and BN-L samples unveiled 57 differential metabolites, and five thereof (involving increased 5-hydroxymaltol, maltol, dipterine, fragransin B2, and decreased adipic acid) were selected as processing-associated markers. These five markers achieved 90 % accuracy in identifying the BN-M samples in the external validation, much better than that gained by seven potential volatile markers recorded by gas chromatography-mass spectrometry. Conclusively, integration of chemometrics and ML reveals the subtle chemical changes during the processing of nutmeg, and it is the first study reporting robust markers that are diagnostic to discriminate between the raw nutmeg and processed products.