Integrating untargeted metabolomics and targeted analysis for not from concentrate and from concentrate orange juices discrimination and authentication

Untargeted metabolomics was used to investigate the differences between orange juices from concentrate or not. Although no distinctive separation could be observed from principal component analysis for all samples, student’s t -test, fold change, orthogonal projection to latent structures discriminant and detection rate analyses were conducted to find the latent potential markers. Subsequently, 91 and 42 potential markers were defined in positive/negative mode. Thirteen of them, l -glutamine, carvone, erucamide, lysoPE (16:0/0:0), oleic acid, α-linolenic acid and 7 tripeptides (reported for the first time in orange juice) were identified with standards. A partial least squares discriminant analysis model, based on the potential markers in positive mode by direct analysis in real time mass spectrometry-quadrupole time of flight (DART-QTOF) with 9 s acquisition time, was constructed and validated with 97% and 95% accuracy for training and test. The model was applied to commercial samples successfully, and an NFC brand was found highly suspicious.