Skip to Main content Skip to Navigation

Food authentication by Proton NMR spectroscopy in combination with chemometric tools

Marion Cuny
Abstract : 1H NMR spectroscopy is widely used as an analytical method in different sectors. Its number of applications in the area of food and beverage authenticity is growing. However, the multivariate analysis of the type of data obtained from an NMR spectrum is still not as developed as in Near Infrared spectroscopy. In this work, different chemometric methods have been applied to 1H NMR data in order to assess the potential of the combined techniques to authenticate food and beverages. First, we have demonstrated that Independent Component Analysis (ICA) was better adapted to the analysis of a 1H NMR spectrum than the commonly used Principal Component Analysis (PCA). Indeed, by its very nature, ICA aims at recovering pure sources from mixed signals. Applied to the spectra, it was shown that it is possible to extract certain component signals, such as the naringin signal from the spectra of a mixture of orange and grapefruit juices. Different pre-treatments were then tested. Data warping has been found to be useful when the data shows variation in chemical shifts. In addition, as the data also varied considerably in intensity along the spectrum, a logarithmic transformation was performed to produce unbiased results when using other chemometric tools such as PCA and ICA. Finally, different approaches were investigated to select variables in the spectrum. The first approach was based on criteria related to the variables themselves, such as the total variance and covariance in the Clustering of Variables (CLV) function. The second type of method involved the selection of contiguous variables to take into account the relation between variables in a signal. Interval_PLS (iPLS) was used as a reference to compare other more recently developed methods: Evolving Windows Zone Selection and Interval-PLS_Cluster. The variable selection techniques used were to highlight known authenticity markers of orange and grapefruit juices: flavonoids hesperidin and naringin that are measured in the standard HPLC method IFU 58. In the case of the balsamic vinegar dataset, the selected zones contained the signal of compounds linked to the product's aging process that differentiate traditional balsamic vinegar from its cheaper, more commonly used counterpart. In the example based on different types of yoghurts, the variable selection procedures focused on 6 certain aroma compounds and solvents used as flavour carriers. These were used to differentiate the type of yoghurts: flavoured, with fruit, with pulp and at different concentrations. Through these different applications, this study has shown the importance of using appropriate tools for spectral analysis tools that take into account the specificity of 1H NMR spectroscopy compared to other spectroscopic methods: variation in intensities along the spectrum, the size of the dataset, and redundant information.
Document type :
Complete list of metadatas

Cited literature [119 references]  Display  Hide  Download
Contributor : Ecole Agroparistech <>
Submitted on : Friday, August 1, 2008 - 8:00:00 AM
Last modification on : Friday, October 23, 2020 - 4:38:06 PM
Long-term archiving on: : Friday, September 10, 2010 - 12:32:24 PM


  • HAL Id : pastel-00003745, version 1



Marion Cuny. Food authentication by Proton NMR spectroscopy in combination with chemometric tools. Chemical Sciences. AgroParisTech, 2008. English. ⟨NNT : 2008AGPT0017⟩. ⟨pastel-00003745⟩



Record views


Files downloads