| Literature DB >> 33103896 |
Juliane Klare1,2, Marc Rurik3, Eric Rottmann1, Anke Bollen1, Oliver Kohlbacher3,4,5,6, Markus Fischer2, Thomas Hackl1,2.
Abstract
Food authenticity concerning the geographical origin becomes increasingly important for consumers, food industries, and food authorities. In this study, nontargeted 1H NMR metabolomics combined with machine learning methodologies was applied to successfully distinguish the geographical origin of 237 samples of white asparagus from Germany, Poland, The Netherlands, Spain, Greece, and Peru. Support vector classification of the geographical origin achieved an accuracy of 91.5% for the entire sample set and 87.8% after undersampling the majority class. Important regions of the spectra could be identified and assigned to potential chemical markers. A subset of samples was compared to isotope-ratio mass spectrometry (IRMS), an established method for the determination of origin of white asparagus in Germany. Here, SVM classification led to accuracies of 79.4% for NMR and 70.9% for IRMS. Finally, the classification of asparagus from different German regions was evaluated, and the influence of year and variety was analyzed.Entities:
Keywords: NMR; asparagus; food fraud; geographical origin; support vector machine
Mesh:
Substances:
Year: 2020 PMID: 33103896 DOI: 10.1021/acs.jafc.0c05642
Source DB: PubMed Journal: J Agric Food Chem ISSN: 0021-8561 Impact factor: 5.279