Description
Black pepper is a valuable commodity susceptible to economically motivated adulterations. This contribution describes a non-targeted method for authentication of black pepper by near infrared spectroscopy (NIR) coupled to least absolute shrinkage and selection operator (LAS-SO). An amount of 116 samples were collected and analyzed. The dataset included authentic black pepper samples (from different Asian and South American regions and four harvesting seasons) and samples spiked with different adulterants, including both endogenous sub-products (pinhead and spent) and exogenous materials (papaya seeds, coriander, plaster, green anise, olive kernel, talc, rice flour, black mustard, green lentil, red beans, chili, garlic and corn flour). The percentage of adulteration ranged between 1.5% and 35%.
The spectra were split in training and test set and then normalized by multiplicative scatter correction (MSC). While the test set was withheld for further testing of the model, the train-ing set was submitted to LASSO with the aim to classify the samples as authentic, exogenous-ly-adulterated and endogenously-adulterated group. LASSO model is used for selection and shrinkage of parameters. The resulting model was cross-validated achieving high overall ac-curacy. The model was tested on the test set achieving an overall accuracy of 94% and very high sensitivity and specificity rates. The model was then validated with a batch of 34 inde-pendent samples and it classified correctly 33/34 samples. The model was also challenged with samples from an inter-laboratory proficiency test. Multiple laboratory verifications of the proposed methodology are ongoing.
Finally, a Shiny web application was set up for direct statistical analysis of the raw data. The NIR user simply uploads the raw data to the app and MSC normalization and interrogation of the LASSO model is performed. The online application facilitates the model testing and ena-bles clear results visualization. This is the first application of LASSO to spectroscopy data.
