Quality assessment of Brazilian olive oils by GC×GC-MS and chemometrics.

Virgin olive oil quality assessment is critical for consumer protection and industry regulation. Fraudulent activities involving virgin olive oil are common worldwide. In Brazil, recent nationwide inspections have revealed recurrent cases of fraud, including the sale of adulterated or mislabeled olive oils. As a result, the Brazilian federal agency has suspended the commercialization of numerous brands and seized large quantities of products that failed to meet the official quality requirements. This study presents an automated analytical workflow for classifying Brazilian virgin olive oils based on their volatile organic compound (VOC) profiles. Headspace solid-phase microextraction (HS-SPME) and comprehensive two-dimensional gas chromatography-mass spectrometry (GC×GC-MS) were used to analyze 215 certified virgin olive oil samples, categorized as either defective (VOO, virgin and LVOO, lampante virgin) or non-defective (EVOO, extra virgin), and 21 undisclosed oils (UNKN). Chemometric modeling using partial least squares-discriminant analysis (PLS-DA) was applied to the resulting data, which comprised 108 recurring VOCs. The PLS-DA model successfully differentiated between the two quality classes (defective and non-defective) with a predictive accuracy of 91% based on the external validation set. Key chemical markers driving the classification were identified. Non-defective oils were characterized by higher levels of C5 and C6 aldehydes and alcohols from the lipoxygenase (LOX) pathway, such as (E)-2-hexenal, associated with positive green and fruity notes. Conversely, defective oils showed higher levels of compounds such as nonanal and acetic acid, linked to rancidity and other off-flavors. Finally, the model was used to predict the quality of 21 previously undisclosed samples. This instrumental approach demonstrates a powerful and reliable alternative for forensic analysis of olive oils, generating models that can be interpreted in sensory terms.