Geographical traceability of cultivated Paris polyphylla var. yunnanensis using ATR-FTMIR spectroscopy with three mathematical algorithms

Abstract

Paris polyphylla has been used by multiple nationalities as a traditional herb medicine to treat diseases in different regions of China. Since Paris quality is influenced by geographical regions, a fast and effective geographical traceability method is necessary. In our study, the geographical origin discrimination of 789 P. yunnanensis species from central, western, northwest, southeast and southwest Yunnan with 3rd to 8th cultivation years was carried out by chemometric of partial least squares discriminant analysis (PLS-DA), random forest (RF) and hierarchical cluster analysis (HCA) combined with attenuated total reflection-Fourier transform mid infrared (ATR-FTMIR). The results indicated that principal component analysis (PCA) was successfully used as an exploration data analysis method to remove outliers. Additionally, classification ability of PLS-DA model, established by variable importance for the projection (VIP), was similar to that of PLS-DA model, which indicated that raw ATR-FTMIR spectra contained abundant redundant information useless for classifying samples. For each model, the range of VIP values of PLS-DA is wider than the range of the important variables of RF, and PLS-DA was more time-efficient and high-ability than RF model. Besides, the geographical origin PLS-DA classification model with all collected samples was used to verify the results with accuracy rates of calibration set and validation set of 98.34% and 93.78%, respectively. HCA verified that the impact of geographical origin on P. yunnanensis characteristics is greater than that of the number of cultivation years. In short, PLS-DA model can assure the accuracy and limit time for identification analysis of cultivated P. yunnanensis quality assessment from different geographical origins.