Evaluation of innovative tools for the rapid and accurate authentication of expensive and highly used food ingredients

The EC published the results of the first coordinated control plan on the authenticity of herbs and spices on the EU market in 2021. 21 EU members, Switzerland and Norway participated. 17% of ground black pepper was found suspicious of adulteration. Here, we evaluated three non-targeted methods for authentication of spices by NIR, GC-HRMS and DART-TIMS-QTOF. Multivariate statistical analysis was performed on the same sample set. The evaluation included authentic black pepper samples from 8 different countries and four harvesting seasons and samples spiked with non-functional material (pinhead and spent) as well as exogenous materials (green lentil, olive kernel, black mustard, sesame, garlic, corn flour, rice flour, chili, papaya). The percentage of adulteration ranged between 5% and 30%.

A method was developed and validated by NIR that achieved high overall accuracy, sensitivity and specificity rates on the test set, the validation set with multiple operators and a proficiency test. One of the most underrated issues of non-targeted methods is the time spent to manually average, normalize and send the data to the statistician for statistical assessment of the acquired data. For this reason, a local web application was developed that allowed the direct interrogation of the statistical model.

Afterwards, a GC-IMS classifier was developed and validated that showed high overall accuracy ≥90% both on the withheld test sets 1 and 2. HS-GC-IMS is characterized by the destruction of the sample as compared to spectroscopy methods. Moreover, the analysis of a single sample by HS-GC-IMS takes about 17 min, which is a longer time as compared with those of DART-MS and NIR spectroscopy. The capability of DART-QTOF-MS with and without TIMS was also evaluated for black pepper authentication. Analysis times were 5 s per sample and therefore significantly shorter than for NIR and GC-IMS analyses. Unsupervised statistical analysis in form of Principal Component Analysis (PCA) revealed a clear discrimination of atypical samples from those authentic. Machine learning classifiers, based on DART-QTOF-MS data, are being built-up and validated. Although the addition of a separation step after ionization using TIMS did not further improve the discrimination, it showed great potential in terms of identifying specific marker compounds for adulterants through cleaner MS/MS spectra and collision cross-section values as an additional identification criterion. Moreover, further challenges of the tools with independent sample set need to be systematically performed in consecutive studies to control the performances of the methods for black pepper authenticity over a longer time frame.