Authentication of Dendrobium Officinale from Similar Species with Infrared and Ultraviolet-Visible Spectroscopies with Data Visualization and Mining. (23rd July 2020)

• Record Type: Journal Article
• Authors: Wang, Ye; Huang, Heng-Yu; Wang, Yuan-Zhong
• Is Part Of: Analytical letters. Volume 53:Number 11(2020)
• Page Start: 1774
• Page End: 1793
• Publication Date: 2020-07-23
• Subjects: Green chemistry -- dendrobium -- data mining -- data visualization -- spectroscopy; Chemistry, Analytic -- Periodicals; Chemistry, Analytic -- Abstracts; 543
• DOI: 10.1080/00032719.2020.1719126
• Languages: English
• ISSNs: 0003-2719

Abstract

Abstract: Dendrobium officinale is utilized as an original plant of Fengdou and Huangcao, which is consumed as a functional and tonic food in daily life owing to its unique characteristics of nourishing stomach. However, there are many similar species named Fengdou for trade in the herbal market while some species are specially used as medical materials in clinics. Therefore, it is necessary to establish a rapid and effective method for controlling their plant origins. In our study, two types of spectra combined with unsupervised and supervised pattern recognition were investigated for authentication of 17 Dendrobium species (170 samples). Three fusion strategies equipped with two data mining methods (contribution variable fusion and principal component combination) were employed to further improve the discrimination performance. The results indicated that spectroscopy was a powerful tool for the authentication of the original plant of Fengdou. Thereinto, ultraviolet-visible spectra after pretreated by the second derivative was beneficial for visualization, especially with the help of t-distributed stochastic neighbor embedding. The contributing variables from two spectroscopies were fused for the establishment of a model based upon partial least squares discrimination analysis, which was beneficial for improving the robustness of the well-established protocol. Finally, these contributions of the variables were calculated by hierarchical cluster analysis for a comprehensive evaluation of their similarity and showed that edible D. officinale was different from the other 16 species. Generally, the data visualization and mining strategy are effective approaches for original plant authentication of Fengdou materials in the herbal market.