Determination of multiple elements in samples of the medicinal plant Marsdenia tenacissima and estimation of geographic origin via pattern recognition techniques.

Multi-element analysis of the medicinal plant Marsdenia tenacissima was used to develop a reliable method of tracing the geographical source of the samples. The concentrations of 27 elements in 128 samples from 4 provinces in China were analyzed by inductively coupled plasma-atomic emission spectroscopy. Pattern recognition techniques, viz. principal component analysis (PCA), cluster analysis (CA), stepwise linear discriminant analysis (SLDA) and k-nearest neighbor analysis (KNN), were used for this purpose. It was verified that 21 elements in the M. tenacissima samples from different regions showed significant differences (P < 0.05). The PCA explained 87.36 % of the variance with the first seven principal component variables, and a score plot produced from the largest three principal components showed that the source area of most samples could be correctly distinguished. The CA showed that samples were separated into three clusters. The SLDA produced an overall correct classification rate of 87.5 % and a cross-validation rate of 85.2 %. The KNN analysis performed ideally, with an average identification rate of 100 % for the training set and 93.33 % for the test set. These results laid the foundation for the application of multi-element analysis combined with pattern recognition techniques for tracing the geographical origin of samples of medicinal plants.