Application of a potentiometric electronic tongue for assessing phenolic and volatile profiles of Arbequina extra virgin olive oils

Abstract

A B S T R A C TThe capability of determining the phenolic and volatile profiles of olive oils is of major relevance since these compounds are known to greatly influence the gustatory and olfactory positive attributes of olive oils. An electronic tongue with multiple linear regression models was used to evaluate both profiles based on olive oils potentiometric data generated during a single assay. The proposed electronic tongue-chemometric procedure enabled the quantification of flavonoids, phenolic acids and phenol alcohols of Arbequina extra-virgin olive oils with a similar accuracy of UPLC-MS (0.93 ± 0.03 ≤ R 2 ≤ 0.98 ± 0.08 for the repeated K-fold cross-validation procedure). Also, it was verified that the potentiometric device should not be applied to evaluate volatile compounds in solution (0.80 ± 0.14 ≤ R 2 ≤ 0.94 ± 0.05 for the repeated K-fold cross-validation procedure),showing a lower accuracy than HS-SPME-GS-MS. The overall satisfactory results showed that electronic tongue could be used as a practical sensing instrument to generate a chemical profile of the compounds known to influence the positive sensory attributes of olive oils.

Application of an electronic tongue as a single-run tool for olive oils’ physicochemical and sensory simultaneous assessment

“…3) confirmed the possibility of using a MRLM-SA-E-tongue model, based on signal profiles of 12 sensors, for predicting the TP contents of olive oils from different olive cultivars (repeated K-fold-CV: R 2 = 0.92 ± 0.07; RMSE = 22 ± 8 mg CAE/kg of olive oil). The quality of the results indicates that the E-tongue-based procedure had, from a statistical point of view, a similar accuracy compared to the HPLC conventional method, confirming the reported successful performance of these type of potentiometric devices for TP evaluation [18,28]. Furthermore, the potentiometric E-tongue predictive performance was comparable to that achieved with voltammetric E-tongues, described in the literature [11,14,15].…”

“…Multiple linear regression (MLR) models were used to estimate and/ or predict the experimental contents of the different physicochemical, chemical and sensory properties of olive oils (evaluated using titration, UV/Vis spectrophotometry, colorimeter, liquid chromatography and/or sensory evaluation by trained panellists), based on the potentiometric E-tongue signal profiles recorded during the analysis of the hydroethanolic extracts. The best number of sensors, with non-collinear potentiometric signals, used for quantitative purpose, was established by applying the simulated annealing (SA) meta-heuristic algorithm, which selection performance was previously demonstrated by the research team for MLRM-SA-E-tongue models [18,20,22,42]. The quality of the multivariate models was assessed through the coefficient of determination (R 2 ) and the root-mean-square error (RMSE), calculated two cross-validation (CV) variants: the leave-one-out CV (LOO-CV) and the repeated K-fold-CV procedures.…”

“…The quality of the multivariate models was assessed through the coefficient of determination (R 2 ) and the root-mean-square error (RMSE), calculated two cross-validation (CV) variants: the leave-one-out CV (LOO-CV) and the repeated K-fold-CV procedures. This latter variant is less prone to overfitting issues since it allows to leave 1/K×100% of the initial data for internal-validation purposes (in this study, K was set equal to 4, enabling that at each run, 25% of the initial data were used for validation purposes, i.e., olive oils produced from 5 to 6 independent centenarian olive trees of a total of 23 trees), being the data split process randomly repeated, which was set equal to 10 in this work [18,20,22]. Finally, the possibility of using the selected E-tongue-MLR-SA models (for both LOO-CV and repeated K-fold-CV procedures) as complementary tools for the analytical conventional assessment of the physicochemical, chemical and sensory data was further checked, as suggested by Roig and Thomas [43,44].…”

“…panellists of the School of Agriculture of the Polytechnic Institute of Bragança (Portugal), following the IOC regulations [4,5] (intensity scale: from 0 (absence of attribute) to 10 (maximum attribute intensity)). g E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:3, S1:4, S1:6 to S1:8, S1:12, S1:13, S2:8, S2:10, S2:15, S2:16 and S2:20. h E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:4, S1:7, S1:10, S1:11, S1:13 to S1:15, S1:20, S2:2, S2:5, S2:7, S2:8 and S2:16. i E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:4, S1:12 to S1:14, S2:2, S2:4, S2:5, S2:7, S2:8, S2:15, S2:16 and S2:20. j E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:5, S1:8, S1:10, S1:13, S1:14, S1:19, S2:3, S2:8 to S2:10, S2:14, S2:16 and S2:20. k E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:6 to S1:9, S1:11, S1:12, S1:20, S2:1, S2:7, S2:8, S2:12, S2:18 and S2:19. l E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:5, S1:9, S1:11 to S1:13, S1:17, S2:2, S2:3, S2:5, S2:8, S2:11, S2:14 and S2:18. m E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:3, S1:5, S1:6, S1:11, S1:12, S1:15, S1:16, S1:19, S2:4, S2:6, S2:9, S2:11 and S2:12. n E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:2, S1:6, S1:7, S1:11, S1:14, S1:15, S1:17, S1:20, S2:11, S2:15, S2:17, S2:18 and S2:20. o E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:1 to S1:4, S1:7, S1:13, S1:18, S1:19, S2:5, S2:10, S2:11, S2:17, S2:19 and S2:20. p E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:1, S1:5, S1:6, S1:9, S1:10, S1:18, S1:20, S2:1, S2:5, S2:8, S2:14, S2:16 and S2:18. q E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:4 to S1:6, S1:12, S1:18, S2:2, S2:4, S2:5, S2:10, S2:13, S2:16, S2:17 and S2:20. r E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:1, S1:3, S1:5, S1:12, S1:13, S2:1 to S2:3, S2:5, S2:7, S2:8, S2:11, S2:18 and S2:20. s E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:3 to S1:6, S1:11, S1:18 to S1:20, S2:1, S2:3, S2:18 and S2:20. t E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:4, S1:5, S1:8, S1:13, S1:19, S2:3, S2:8 to S2:11, S2:16, S2:17 and S2:20. u E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:4, S1:8, S1:10, S1:12, S1:20, S2:1, S2:3, S2:4, S2:11, S2:14 and S2: 18. v E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:3, S1:5, S1:7, S1:10, S1:11, S1:16, S2:8, S2:10, S2:14, S2:15 and S2:19. w E-tongue sensor signals included in the E-tongue-MLR-SA model: S1:5, S1:8, S1:10, S1:11, S1:13 to S1:15, S1:18, S2:2, S2:7, S2:9, S2:14 and S2:18.…”

“…This evaluation is of major relevance since polar phenols are related with beneficial health and sensorial properties, while also contributing for olive oil resistance to oxidation. Previous works have reported the successful use of potentiometric E-tongues for assessing the TP levels in Arbequina olive oils [18] and on vegetable oils [28]. So, the capability of the novel potentiometric E-tongue device was also studied for this specific application.…”