Comprehensive quality assessment of Dendrubium officinale using ATR-FTIR spectroscopy combined with random forest and support vector machine regression

Abstract

Dendrobium officinale, as a tonic herb, has attracted more and more consumers to consume in daily life. In order to protect the wild resource, the herb has made great progress though cultivation in vitro. However, the quality is fluctuated in Chinese herbal medicine market due to influence such as cultivated areas and harvesting period. Therefore, the herbal samples from different cultivated locations were evaluated with high-performance liquid chromatography with diode array detector (HPLC-DAD) in terms of two chemical components, quercetin and erianin. In addition, two markers in leaf and stem also were used for support vector machine regression (SVMR) prediction. Samples from different harvesting periods were also classified using attenuated total reflectance mid-infrared spectroscopy coupled with random forest model. The results indicated that Pu'er and Menghai in Yunnan Province were suitable places for the herb cultivation and the leaf of the herb was also an exploitable resource just in light of the content of two components. What's more, combination of suitable spectra pretreatment and grid search method efficiently improved the prediction performance of the regression model. The results of random forest model indicated that important variables combination between stem and leaf was an effective tool to predict the harvesting time of the herb with 94.44% accuracy in calibration set and 97.92% classification correct rate in validation set. The results of combination were better than the models using individual stem and leaf spectra. In addition, the suitable harvesting time (December) could be classified efficiently. Our study provides a reference for quality control of raw materials from D. officinale herb.