Classification of Rosa roxburghii Tratt from different geographical origins using non-targeted HPLC-UV-FLD fingerprints and chemometrics

In this study, a novel non-targeted strategy based on high-performance liquid chromatography with ultraviolet detection (HPLC-UV) and fluorescence detection (HPLC-FLD) was first proposed for the classification of Rosa roxburghii Tratt (RRT) from eight geographical origins in Guizhou, China. HPLC-UV and HPLC-FLD fingerprints were simultaneously recorded by an HPLC-UV-FLD instrument. Then, fingerprint data were processed with low-level data fusion and variable reduction strategies before chemometric analysis. Based on different signal types and data orders of the resulting fingerprints, four strategies for RRT classification were proposed and compared. In the first three strategies, three supervised classification methods including 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 different kinds of first-order fingerprints (HPLC-UV, HPLC-FLD and HPLC-UV-FLD), respectively. Moreover, N-way partial least squares-discriminant analysis (NPLS-DA) discriminant model was established based on the second-order fingerprints acquired by HPLC-FLD. By comparison, the best result was obtained by PLS-DA based on first-order HPLC-UV-FLD fused fingerprints, the correct classification rates (CCRs) of cross-validation, training set and test set were 98.8%, 100% and 96.9%, respectively. Non-targeted chromatographic fingerprints were used to solve the problem of RRT classification for the first time.