Purpose: The aim of this work is to develop classification models for accessing freshness (e.g., microbiological and organoleptic parameters) in beef filet samples using metabolomics and fingerprinting with conventional (HPLC) and non-destructive (FTIR, e-nose) instrumentation using support vector machines (SVMs).
Methods: In the framework of SYMBIOSIS-EU project, the shelf life of beef filets stored aerobically at 0, 5, 10, 15 and 20°C was investigated. The microbial association of meat and the temporal biochemical changes were monitored. Microbiological analyses, including total viable counts, pseudomonads, Brochothrix thermosphacta, lactic acid bacteria, and Enterobacteriaceae, were undertaken, while in parallel sensory assessment, pH measurement, HPLC analysis of the organic acid profiles, FT-IR, and eNose measurements were recorded and the data were analyzed.
Results: The data derived from HPLC, e-nose and the fingerprint from FTIR were used to develop two sets of SVM models: (i) based on individual and (ii) combined datasets using all possible pairwise combinations, in order to assess the effect of combined data in improving the prediction performance of the developed models. Model performance was assessed using independent subsets. The models were then optimized to achieve the best classification accuracy for specific data types or data type combinations.
Significance: The significance of these findings relies on the fact that such approaches can provide reliable indication of the quality status of meat in retail regardless of whose perspective you take, i.e., that of the consumer, the industry, the inspection authority, or the scientist.