Method M4B
Untargeted approach for the analysis of virgin olive oil volatile fingerprint by SPME-GC-MS and chemometrics
The method describes a fit for purpose screening strategy to discriminate virgin olive oils (VOO) into the ‘Extra virgin’, ‘Virgin’ or ‘Lampante’ olive oil quality grades; aiming to support the sensory panels by reducing its workload to boundary samples and therefore improving its performance1. It is based on a hierarchical classification model built with the volatile fingerprint of VOO samples, which are obtained by solid-phase microextraction (SPME) and subsequent separation of analytes by gas chromatography coupled with a mass spectrometry (MS) detector.
Consensus about the procedure to validate fingerprinting analytical methods, like the present one, is still lacking and precise guidelines have not been set2. Therefore, we addressed a peer validation study that focuses on evaluating the raw signal provided by the participants, understanding the variability and the sources of error that will help to build a robust classification model in the future. The main constraint of the fingerprinting analytical method is in providing a repeatable and reproducible chromatographic signal with enough sensitivity to collect the valuable information from the samples, thus intra and inter-lab repeatability are assessed for the 4 participants.
First, the intra-day repeatability of each participant was evaluated by calculating the Relative Standard Deviation (RSD) at each data point – excluding noise - of the total ion chromatogram (TIC). All participants achieved an RSD < 20% in more than 95% of the signals. Moreover, an inverse relationship between RSD% values and signal intensity was observed when plotting them, agreeing with the Horwitz equation3.
Then, inter-lab reproducibility was checked by principal component analysis (PCA), after signal normalization and alignment to solve the retention time shifting typical of chromatographic data. PCA showed a satisfactory clustering of each pre-trial sample, indicating a precise analytical outcome (Figure 1). Moreover, pooled quality control (QC) cluster is located near to the QC analysed previously by the developers, indicating a satisfactory reproducibility. The closeness of the pooled QC cluster to the Extra Virgin and Lampante samples used to prepare the pooled QC, suggests a stronger weight of the variables related to these sensory profiles.
Under this scope, the results obtained in the pre-trial were promising; the type and the extent of the variability between the signals have been assessed and some sources of error have been identified. The limited number of participants and samples hindered the potential to better capture the instrumental and analytical variability.
The trial proper phase has been carried out in Autumn 2020.
1. B. Quintanilla-Casas, et al. Virgin Olive Oil Volatile Fingerprint and Chemometrics: Towards an Instrumental Screening Tool to Grade the Sensory Quality. LWT - Food Science and Technology 2020, 121, 108936.
2. J. Riedl, S. Esslinger, C. Fauhl-Hassek. Review of validation and reporting of non-targeted fingerprinting approaches for food authentication. Analytica Chimica Acta (2015) 885:17-32.
3. Horwitz, W. Evaluation of Analytical Methods Used for Regulation of Foods and Drugs. Analytical Chemistry 1982, 54, 67–76.
Figure 1. Principal Component Analysis (PCA) on TIC relative intensities (normalized, non-scaled) of: pre-trial samples analysed by the participants (10, OLPT_M4b_010 pooled QC; 11, OLPT_M4b_011- Lampante Oil; samples analysed by the reference lab during in-house validation (QC_VAL, pooled QC; EV_RSs, Extra Virgin Reference Samples, L_RSs, Lampante Reference Samples).
Note: PCA characteristics: 2PC, 89.4% of explained variance.
For the University of Barcelona developing team:
Stefania Vichi and Alba Tres