Meat is a highly consumed product widely susceptible to fraudulent practices. Among the authenticity issues that have begun to be considered by society are meat origin (geographical indication), production practices (organic), and ethical and religious aspects (animal welfare, Halal and Kosher foods, etc.). Although genetic analyses can resolve authentication aspects related to animal species, the factors discussed above cannot be solved genetically. Thus, metabolomics emerges as a strategy that could solve these cases of food fraud since it focuses on analysis of the metabolites present in meat, which will depend on external factors such as stress, diet, production area, etc.

The capacity of a non-targeted HPLC-UV metabolomic method for the classification and authentication of meat products was evaluated. A total of 200 meat samples were analyzed, including white meat (chicken, turkey, duck, and quail) and red meat (beef, lamb, pork, and rabbit). In the case of the lamb samples, they were analyzed from two different geographical origins (Catalonia and Aragon), and the chicken samples came from two different production systems (organic and conventional). Simple extraction of the metabolites was carried out (ultrasound with water), and the extracts were analyzed with the proposed non-targeted HPLC-UV fingerprinting method, using chromatographic fingerprints as the chemical descriptors for chemometric analysis by PLS-DA. The capacity of the proposed methodology to classify and authenticate the eight meat typologies using PLS-DA was excellent, with sensitivity and specificity values higher than 95.0% and 95.7%, respectively, and classification errors, in most cases, below 0.4% (2% only in the quail samples). The classification using PLS-DA was perfect (100% classification rate) when considering the white or red meat samples independently. This method was also excellent for authenticating the geographical origin of the lamb samples and the production system (organic vs. conventional) in the chicken samples, with PLS-DA classification errors lower than 2.3% and 0%, respectively.