Non-targeted NMR-based approach has received great attention as a rapid method for food product authenticity assessment. The availability of a database containing many comparable NMR spectra produced by different spectrometers is crucial to develop functional classifiers able to discriminate rapidly the commodity class of a given food product. Nevertheless, variability in spectrometer features may hamper the production of comparable spectra due to inherent variations in signal resolution. In this paper, we report on the development of a class-discrimination model for grape juice authentication by application of non-targeted NMR spectroscopy. Different approaches for the pre-treatment of data will be described along with details about the model validation. The developed model performed excellently (95.4-100% correct predictions) even when it was tested against 650 spectra produced by 65 spectrometers with different configurations (magnetic field strength, manufacturer, age). This study may boost the use of non-targeted NMR methods for food control.
Keywords: Chemometric analysis; Fingerprinting; Food authenticity; Grape juice; Interlaboratory variability; Metabolite profiling; Method validation; Non-targeted NMR-based metabolomics approach.
Copyright © 2020 Elsevier Ltd. All rights reserved.