Improving the estimations of fatty acids in several Andalusian PDO olive oils from NMR spectral data

ABSTRACT

The aim of this paper is to determine the fatty acid profile of diverse Andalusian extra-virgin olive oils from different protected designations of origin (PDO). The available data for the statistical multivariate analysis have been obtained from gas chromatography (GC, used as classical reference analytical technique) and nuclear magnetic resonance (NMR) spectroscopy : ${}^1\mathrm{H}$-NMR and ${}^{13}\mathrm{C}$-NMR (in the carbonyl, C-16 y aliphatic carbon regions). The diverse percentages of fatty acids approximated by the above-mentioned chemical procedures are summarized by using a statistical treatment, which presents a some weighted averages to obtain the closest fatty acid profile to the one provided by the GC reference technique, with weights being inversely proportional to some measures of the calibration errors. Besides, the work shows that the PDO of an olive oil conditions the NMR region (${}^1\mathrm{H}$-NMR or carbonyl, C-16 or aliphatic ${}^{13}\mathrm{C}$-NMR) which provides the best estimation of each type of fatty acid. Finally, procedures of cross-validation are implemented in order to generalize the previous results.

KEYWORDS:

• Extra-virgin olive oil
• ${}^1\mathrm{H}$ and ${}^{13}\mathrm{C}$ NMR
• protected designation of origin
• multivariate statistics
• cross-validation

AMS:

• 62H25
• 62J05
• 62Q99

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

M. I. Sánchez-Rodríguez http://orcid.org/0000-0003-0657-7823

E. M. Sánchez-López http://orcid.org/0000-0002-4885-6065

A. Marinas http://orcid.org/0000-0002-2693-0711

J. M. Caridad http://orcid.org/0000-0003-4558-6618

F. J. Urbano http://orcid.org/0000-0002-3489-1601

Notes

1. As indicated by Berrueta et al. [3], the ideal situation in the evaluation of the predictive capability of a prediction is when there are enough data available to create a validation set (test data set) completely independent from the calibration set (learning data set). But cost or time constraints in the acquisition of data processes usually complicate the availability of independent validation subsets. Therefore, the initial sample is subdivided in calibration and validation subsets.

2. The chemometric applications can be developed using different software. Some packages of statistical or mathematical analysis have implemented the principal techniques usual in Chemometrics, such as the PLS Toolbox of MatLab -from MathWorks-, Microsoft Excel, PASW Statistics -formerly SPSS, currently belonging to IBM-, UNSCRAMBLER -from CAMO-, or the free package‘pls’  in R-Project.