Non-targeted HPLC–FLD fingerprinting for the classification, authentication, and fraud quantitation of Guizhou paprika by chemometrics

In this work, non-targeted strategies based on high-performance liquid chromatography with fluorescence detection (HPLC–FLD) fingerprints were proposed as chemical markers to address the classification, authentication and fraud quantitation of paprika from six geographical origins in Guizhou, China. Partial least squares-discriminant analysis (PLS-DA), principal component analysis-linear discriminant analysis (PCA-LDA) and random forest (RF) were used to build discriminant models using first-order fingerprints, while n-way partial least squares-discriminant analysis (NPLS-DA) was used to build the discriminant model using second-order fingerprints. Given the large differences in the paprika fingerprints from different origins, all methods achieved satisfactory classification results, with the recognition rate of training set and prediction set reaching 100%. Moreover, the fingerprints were also proposed to detect and quantify three paprika geographical origin blend cases by partial least squares (PLS), random forest (RF) and N-way partial least squares (N-PLS) regression algorithms. When first-order fingerprints were applied, PLS showed much better results than RF for all adulteration cases studied. Improved performances were observed with N-PLS using second-order fingerprints, exhibiting similar calibration errors and lower prediction errors (≤ 2.53%) in comparison to PLS using first-order fingerprints.