Integrating metabolomic data from multiple analytical platforms for a comprehensive characterisation of lemon essential oils

Abstract

Citrus cold pressed oils are of great importance to the flavour and fragrance industry. Because of their high added value, careful attention must be paid to ensure the oils’ genuineness and authenticity. Characterising their chemical complexity in a holistic perspective constitutes a potent way to relate specific compounds to the organoleptic properties of interest and to assess their quality. In this context, a complete characterisation using untargeted metabolomics represents an analytical challenge. The present study takes advantage of multiblock data modelling to integrate heterogeneous signals collected from GC-FID, 1H-NMR, UHPLC-TOF/MS- (negative mode) and UHPLC-TOF/MS+ (positive mode) platforms to obtain a complete characterisation of 64 samples of cold pressed lemon oil (CPLO). Two statistical approaches (MB-PLS-DA and Consensus OPLS-DA) were used to classify the samples according to their extraction processes [i.e. Sfumatrice, Food Machinery Corporation Inline Extractor (FMC), Brown Oil Extractor (BOE), Pelatrice, or mixed FMC + Pelatrice]. Furthermore, the multiblock strategy allows links between variables from different analytical methods to be drawn easily and facilitates the identification of compounds. Because citrus oils extracted using the Sfumatrice process constitute the reference quality from an organoleptic point of view, these samples were characterised thoroughly and are reported to contain less fatty acids but more sesquiterpenes and furocoumarins compared with products obtained using other extraction processes

Combined Nuclear Magnetic Resonance Spectroscopy and Mass Spectrometry Approaches for Metabolomics

“…These models both performed better than the individual supervised models, with a major preference for the OSC-PLS-DA which resulted in a lowest misclassification score due to the removal of confounding factors following the orthogonal signal correction . Another study, which aimed to differentiate the extraction protocols of cold-pressed lemon oil, supervised modeling analyses, namely, MB-PLS-DA and consensus (C)-OPLS-DA, elegantly showed the benefits of using orthogonal projection to improve the separation between samples . In this illustration, a low-data level fusion of untargeted data sets obtained by 1 H NMR, GC-FID, and LC-MS in positive and negative ionization modes was used (Figure ).…”

“…90 In this illustration, a low-data level fusion of untargeted data sets obtained by 1 H NMR, GC-FID and LC-MS in positive and negative ionization modes was used (Figure 5). 90 The supervised analyses then showed a much better separation, and thus interpretability when it came to the C-OPLS-DA compared to the MB-PLS-DA, although their predictive performance was similar. It is important to note that even if the data matrix resulting from the fusion of these four data sets was extensive, high-level data fusion successfully discriminated the same samples according to their geographical origins in another study, 91 but not according to their extraction processes.…”

“…89 Another study, which aimed to differentiate the extraction protocols of cold-pressed lemon oil, supervised modeling analyses, namely, MB-PLS-DA and consensus (C)-OPLS-DA, elegantly showed the benefits of using orthogonal projection to improve the separation between samples. 90 In this illustration, a low-data level fusion of untargeted data sets obtained by 1 H NMR, GC-FID, and LC-MS in positive and negative ionization modes was used (Figure 5). 90 The supervised analyses then showed a much better separation, and thus interpretability, when it came to the C-OPLS-DA compared to the MB-PLS-DA, although their predictive performance was similar.…”

Chemometric Strategies for Spectroscopy-Based Food Authentication

“…It is also often used in combination with LDA for addressing classification problems-MB-PLS is here known as MB-PLS-LDA or MB-PLSDA. So far, many applications of MB-PLS have been reported in the field of food analysis (see Table 3); it has been resorted to (i) for the investigation of sensory parameters of different nature and of their relationships with technological properties of cheese and bread samples [189,190], (ii) for the prediction of meat spoilage time, wine ageing time and crude protein and moisture content in soybean flour by MIR and NIR spectroscopy [191][192][193], (iii) for the discrimination of botanical varieties of extra virgin olive oil, lemon essential oils and wines of different geographical origin by MIR, NIR and Raman spectroscopy [194][195][196], and (iv) for distinguishing added-value from low-quality products [132].…”

Statistical Integration of ‘Omics Data Increases Biological Knowledge Extracted from Metabolomics Data: Application to Intestinal Exposure to the Mycotoxin Deoxynivalenol

“…This ‘omics data fusion allows us to handle a biological system as a whole to gain insight into its complex functioning through the identification of more informative models [ 11 ]. The main statistical methods used to integrate several ‘omics datasets are the penalized canonical correlation analysis (CCA) [ 12 , 13 ] and self-organizing map (SOM) [ 14 , 15 ], used to compute correlations between blocks, and the multi-block generalizations of partial least squares (PLS) to fit a model between a biological factor and ‘omics profiles [ 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 ].…”