Purpose: The aim of this work was to evaluate the efficiency of Headspace Solid Phase Microextraction Gas Chromatography-Mass Spectrometry in tandem with bioinformatics to discriminate between beef and pork minced meat samples and quantify the microbial load.
Methods: Beef and pork minced meat samples were provided by a meat plant in Athens, analyzed microbiologically (Total Viable Counts, Pseudomonas spp., Brochothrix thermosphacta, lactic acid bacteria, and Enterobacteriaceae), and subjected to HS-SPME-GC/MS. In total, 400 chromatograms corresponding to microbiological counts were collected. The dataset was divided 70/30 for model calibration and validation, respectively. The variables (compounds) that were considered important, through Partial Least Squares Discriminant Analysis (PLS-DA) using the bw regression coefficients, were further used to build PLS-Regression (PLS-R) models for each meat category independently. PLS-DA models were evaluated in terms of sensitivity and overall correct classification, while the performance of PLS-R models was assessed using the bias and accuracy factors, RMSE, and relative error (%).
Results: Meat volatilome for the two types of mince showed both qualitative and quantitative differences. Specifically, PLS-DA showed 100% correct classification for both meat categories. Moreover, for PLS-R models, bias and accuracy factors in both cases were close to 1 for both calibration and validation datasets. Considering RMSE, it was calculated close to 0.5 for both meat categories for the training dataset, whereas for validation it was 0.74 and 0.61 for beef and pork mince, respectively. Lastly, 92% of the predictions fell within +/-20% of relative error.
Significance: In conclusion, this study showed the ability of Gas Chromatography-Mass Spectrometry to mange off-line the daily meat production in qualitative and quantitative terms.