Instrumental GC-MS analysis of virgin olive oils already subjected to sensory evaluation

The aroma plays an important role in the olive oil consumer preference and it is one of the parameters used to classify olive oils. The oils of lower quality have an aroma very different rather than that of an extra virgin olive oil, due to the presence of metabolic pathways different from the Lipoxygenase (LOX) one. Depending on the relevant pathway, different odorants are produced giving rise to unpleasant sensory perception whose intensity is related to the amounts of some aroma components. The sensory evaluation, also called “panel test” is the only normed method to assess the quality of the oils relying on their aroma, but this procedure, although carried out by a trained assessor, has some drawbacks. The use of analytical techniques consists in an objective approach, able to identify and quantify the odorants in the volatile fraction of both extra virgin and virgin oils. In this work, 77 olive oils were analyzed; 21 were extra virgin while 56 were virgin olive oils characterized by different sensory defects with different intensities. SPME-GC-MS techniques and the “Find by Chromatogram Deconvolution” algorithm were applied, in order to extract the most compounds as possible. The results obtained were subjected to some statistical analysis, from the simple Principal Component Analysis (PCA) to the more complex Partial Least Square (PLS) regression, to find some correlations between sensory evaluation and chemical composition, with the final aim to develop a method suitable to verify the results of the panel test. The PCA was not so useful to reach the goal, so the PLS regression was applied. The models obtained highlighted the compounds characterizing the defected samples analyzed, each one with a specific importance. The models developed have been composed by a high number of variables because, instead to consider the compounds concentration, the variables subjected to this analysis have been the chromatographic signal detected at each time of the analysis. To simplify, only the relevant variables were taken into account and some relations between the specific compound content and the median of the defects have been found