Summary: | <sup>1</sup>H NMR spectroscopy, in combination with chemometric methods, was used to analyze the methanol/acetonitrile (1:1) extract of walnut (<i>Juglans Regia</i> L.) regarding the geographical origin of 128 authentic samples from different countries (France, Germany, China) and harvest years (2016–2019). Due to the large number of different metabolites within the acetonitrile/methanol extract, the one-dimensional (1D) <sup>1</sup>H NOESY (nuclear Overhauser effect spectroscopy) spectra suffer from strongly overlapping signals. The identification of specific metabolites and statistical analysis are complicated. The use of pure shift <sup>1</sup>H NMR spectra such as PSYCHE (pure shift yielded by chirp excitation) or two-dimensional ASAP-HSQC (acceleration by sharing adjacent polarization-heteronuclear single quantum correlation) spectra for multivariate analysis to determine the geographical origin of foods may be a promising method. Different types of NMR spectra (1D <sup>1</sup>H NOESY, PSYCHE, and ASAP-HSQC) were acquired for each of the 128 walnut samples and the results of the statistical analysis were compared. A support vector machine classifier was applied for differentiation of samples from Germany/China, France/Germany, and France/China. The models obtained by conduction of a repeated nested cross-validation showed accuracies from 58.9% (±1.3%) to 95.9% (±0.8%). The potential of the <sup>1</sup>H-<sup>13</sup>C HSQC as a 2D NMR experiment for metabolomics studies was shown.
|