Raman spectroscopy and discriminant analysis applied to the detection of frauds in bovine meat by the addition of salts and carrageenan

Abstract In the last years, there has been an important and growing concern about food authentication due to the increasing number of occurrences of new types of food frauds. Recently, some frauds have been reported describing the injection of non-meat ingredients, such as salts and polysaccharide compounds, into bovine meat in natura, aiming at increasing its water holding capacity (WHC) and obtaining economic fraudulent gains. Thus, this paper developed a simple and rapid analytical method based on a multivariate supervised classification model (partial least squares discriminant analysis, PLS-DA) and Raman spectroscopy for tackling this problem. Sixteen vacuum-packed pieces of the same cut, eye of the round (semitendinosus), of approximately 4 kg were obtained from different origins. According to an experimental design, each piece was divided into 11 parts, providing control and adulterated samples. Single, binary and ternary mixtures of adulterated samples were prepared by injecting NaCl, sodium tripolyphosphate and carrageenan in the meat pieces. A total of 165 samples were produced (54 controls and 111 adulterated) and their purges, the exudated liquid extracted from the meat after thawing, were obtained. Raman spectra of these purges were recorded between 1800 and 700 cm−1. The whole data set was split into 112 samples for the training set and 53 for the test set. The best PLS-DA model was built with 4 latent variables and successfully discriminated adulterated samples at relatively small rates of false negative and false positive results, which varied from 8.0 to 11.7%. As an additional validation step, confidence intervals were calculated by bootstrap algorithm.