Main content area

Comparing two metabolic profiling approaches (liquid chromatography and gas chromatography coupled to mass spectrometry) for extra-virgin olive oil phenolic compounds analysis: A botanical classification perspective A

Bajoub, Aadil, Pacchiarotta, Tiziana, Hurtado-Fernández, Elena, Olmo-García, Lucía, García-Villalba, Rocío, Fernández-Gutiérrez, Alberto, Mayboroda, Oleg A., Carrasco-Pancorbo, Alegría
Journal of chromatography 2016 v.1428 pp. 267-279
bioactive properties, chemometrics, cultivars, extra-virgin olive oil, gas chromatography-mass spectrometry, liquid chromatography, metabolomics, models, phenolic compounds, principal component analysis
Over the last decades, the phenolic compounds from virgin olive oil (VOO) have become the subject of intensive research because of their biological activities and their influence on some of the most relevant attributes of this interesting matrix. Developing metabolic profiling approaches to determine them in monovarietal virgin olive oils could help to gain a deeper insight into olive oil phenolic compounds composition as well as to promote their use for botanical origin tracing purposes. To this end, two approaches were comparatively investigated (LC–ESI–TOF MS and GC–APCI–TOF MS) to evaluate their capacity to properly classify 25 olive oil samples belonging to five different varieties (Arbequina, Cornicabra, Hojiblanca, Frantoio and Picual), using the entire chromatographic phenolic profiles combined to chemometrics (principal component analysis (PCA) and partial least square-discriminant analysis (PLS–DA)). The application of PCA to LC–MS and GC–MS data showed the natural clustering of the samples, seeing that 2 varieties were dominating the models (Arbequina and Frantoio), suppressing any possible discrimination among the other cultivars. Afterwards, PLS–DA was used to build four different efficient predictive models for varietal classification of the samples under study. The varietal markers pointed out by each platform were compared. In general, with the exception of one GC–MS model, all exhibited proper quality parameters. The models constructed by using the LC–MS data demonstrated superior classification ability.