Geographical provenancing of purple grape juices from different farming systems by proton transfer reaction mass spectrometry using supervised statistical techniques

BACKGROUND: Organic, biodynamic and conventional purple grape juices (PGJ; n = 79) produced in Brazil and Europe were characterized by volatile organic compounds (m/z 20-160) measured by proton transfer reaction mass spectrometry (PTR-MS), and classification models were built using supervised statistical techniques. RESULTS: k-Nearest neighbours and soft independent modelling of class analogy (SIMCA) models discriminated adequately the Brazilian from European PGJ (overall efficiency of 81% and 87%, respectively). Partial least squares discriminant analysis (PLSDA) classified 100% European and 96% Brazilian PGJ. Similarly, when samples were grouped as either conventional or organic/biodynamic, the PLSDA model classified 81% conventional and 83% organic/biodynamic juices. Intraregional PLSDA models (juices produced in the same region - either Europe or Brazil) were developed and were deemed accurate in discriminating Brazilian organic from conventional PGJ (81% efficiency), as well as European conventional from organic/biodynamic PGJ (94% efficiency). CONCLUSIONS: PGJ from Brazil and Europe, as well as conventional and organic/biodynamic PGJ, were distinguished with high efficiency, but no statistical model was able to differentiate organic and biodynamic grape juices. These data support the hypothesis that no clear distinction between organic and biodynamic grape juices can be made with respect to volatile organic compounds.