Virgin olive oil volatile fingerprint and chemometrics
Virgin olive oils are currently graded into the commercial categories according to chemical-physical parameters as well as sensory assessment, following official methods (EC 2568/91 and its amendments). In fact, the latter analysis still presents some drawbacks despite trained sensory panels must perform the evaluation according to the IOC method (IOC/T.20/Doc. 15/Rev. 10 2018). In particular, the high number of samples that the panels analyse every crop season and the fact that it is a time-consuming methodology, may harm the robustness of the results due to the low possibility of performing repetitions.
For these reasons, the present work aimed to provide support to the sensory panels, lightning their workload and improving their performance, by developing an instrumental screening tool that assigns virgin olive oil samples to a given commercial category. It consists of a sequential classification model was built with the volatile fingerprint of 176 virgin olive oils already graded by six different sensory panels into extra virgin (EV), virgin (V) and lampante (L) categories. Volatile fraction was chosen for this purpose due to the high correlation with virgin olive oil sensory attributes and, besides, its analysis involves an automatable solvent-free technique (HS-SPME-GC-MS) which is and requires low amount of sample. Furthermore, the fingerprinting approach allowed us to include the whole analytical information, since it uses the raw signal, in less time compared to a traditional targeted approach where identification and integration steps are needed.
Therefore, two sequential classification models were developed and validated: the first discriminated EV and non-EV samples; the second classified the latter into V or L categories. The novelty of this strategy is that an uncertainty range was set for each classification model, so samples were directly classified into one category or were identified as boundary; depending on whether their prediction result is in or out of the set range. Thus, sensory panels would assess only the boundary samples (17.8% of the total sampling), while the rest would be classified correctly by the instrumental screening tool (89% of the classified samples). Indeed, considering that the sensory assessment by the panels is the reference method, a 91% of reliable assignment would be achieved by combining both approaches.
Quintanilla-Casas, B.; Bustamante, J.; Guardiola, F.; García-González, D.L.; Barbieri, S.; Bendini, A.; Gallina Toschi, T.; Vichi, S.; Tres, A. Virgin olive oil volatile fingerprint and chemometrics: Towards an instrumental screening tool to grade the sensory quality. (2020). LWT-Food Science and Technology, 121, 108936.