Multivariate analysis for organic milk authentication

Abstract

To differentiate organic milk (OM) from conventional milk (CM), an orthogonal projection to latent structure-discriminant analysis (OPLS-DA) model was constructed using δ13C, δ15N, δ18O, 51 elements and 35 fatty acids (FAs) as the variables. So far, most reported studies barely use three or more types of variables, but more variables could avoid one-sidedness and get stabler models. Our multivariate model combines geographical and nutritional parameters and displays better explanatory and predictive abilities (R2X = 0.647, R2Y = 0.962 and Q2 = 0.821) than models based on fewer variables for differentiating OM and CM. In particular, δ15N, Se, δ13C, Eu, K and α-Linolenic acid (ALA) are found to be critical parameters for the discrimination of OM. These results show that the multivariate model based on stable isotopes, elements and FAs can be used to identify OM, and can potentially expand the global databases for quality and authenticity of milk.