Multivariate approach for the authentication of vanilla using infrared and Raman spectroscopy

Many different versions of vanilla extracts exist in the market in a variety of origins, purity levels and composition with little effective regulation. In this study, vanilla is authenticated both in terms of purity and geographical origin applying a multivariate approach using near infrared (NIR), mid infrared (MIR) and Raman spectroscopy following a complex experimental design. Partial least squares-discriminant analysis (PLS-DA) was applied to the spectral data to produce qualitative models. The prediction accuracy of the models was externally validated from the specific success/error contingencies. The results showed that MIR and Raman are reliable for authenticating vanilla in terms of purity, obtaining sensitivity, specificity, precision, and efficiency values equal to 1.00, and Raman is especially suitable for indicating the geographical origin of vanilla extracts, achieving performance metrics around 0.9.