Abstract
In this paper, a remarkably precise, simple, and objective definition of monofloral and polyfloral honey based on NMR metabolomics is proposed. The spectra of organic extracts of 983 samples of 16 botanical origins were used to derive one-versus-all OPLS-DA classification models. The predictive components of the statistical models reveal not only the principal but also the secondary floral origins present in a sample of honey, a novel feature with respect to the methods present in the literature that are able to confirm the authenticity of monofloral honeys but not to characterize a mixture of honey types. This result descends from the peculiar features of the chloroform spectra that show diagnostic resonances for almost each botanical origin, making these NMR spectra suitable fingerprints. The reliability of the method was tested with an additional 120 samples, and the class assignments were compared with those obtained by traditional analysis. The two approaches are in excellent agreement in identifying the floral species present in honeys and in the botanical classification. Therefore, this NMR method may prove to be a valid solution to the huge limitations of traditional classification, which is very demanding and complex.