Discrimination and Geographical Origin Prediction of Cynomorium songaricum Rupr. from Different Growing Areas in China by an Electronic Tongue

Cynomorium songaricum Rupr. is a well-known and widespread plant in China. It has very high medicinal values in many aspects. The study aimed at discriminating and predicting C. songaricum from major growing areas in China. An electronic tongue was used to analyze C. songaricum based on flavor. Discrimination was achieved by principal component analysis and linear discriminant analysis. Moreover, a prediction model was established, and C. songaricum was classified by geographical origins with 100% degree of accuracy. Therefore, the identification method presented will be helpful for further study of C. songaricum.

In our previous work, we presented that the chemical constituents of C. songaricum from different producing areas vary, thus affecting the quality of the plant [2]. Genuine medicinal herb, which means Daodi yaocai in Chinese, is a unique definition in traditional Chinese medicine. Medicinal herbs growing in a specific place exhibit high quality [10]. Currently, chromatographic herbal fingerprints have become one of the most applied quality control tools for similarity analyses of herbal medicines [11]. However, it costs a relatively long time.
us, a more convenient way for identification and quality control of herbs is needed.
Electronic tongues are analytical systems formed from an array of electrochemical sensors combined with dataprocessing tools intended to interpret electrochemical signals. Similar to human receptors, the sensors of an electronic tongue undergo a series of reactions. While the generated reactions differ from one another, the information acquired from each sensor is complementary. en, the results combined by the sensors generate a unique fingerprint that can reflect the macroscopic characteristics of samples. In biological mechanisms, gustatory signals are transducted by brain nerves in the form of electric signals. Electronic tongue sensors approach flavors similarly, given that electric signals are generated with potentiometric variations. e perception and recognition of taste quality are based on the recognition or building of activated sensory nerve patterns in the brain and the gustation fingerprint of a product. is step is accomplished by the statistical software of the electronic tongue that can translate sensor data into taste patterns [12][13][14][15][16][17][18][19]. In the recent years, electronic tongues have been commonly used to analyze food and beverages, given their advantages of short response time, strong objectiveness, human safety, and repeatability [20]. As for some herbs like C. songaricum, they have different tastes and flavors according to different places of origin and are ready to eat. Based on this, the simplicity and convenience of electronic tongues could be used in the analysis of the herbs.
In this work, we first developed a method to discriminate and predict the geographical origin of C. songaricum from different growing areas in China by using an electronic tongue. Pattern recognition techniques, including principal component analysis (PCA) and linear discriminant analysis (LDA), were used for data analysis in this research. In addition, this study provided a simple approach for identifying the geographical origins of C. songaricum, and the acquired information can be used for evaluating the quality of C. songaricum growing in China.

Instrument.
e electronic tongue system (taste sensing system Astree II, France) consists of a reference electrode and seven liquid sensors (ZZ, JE, BB, CA, GA, HA, and JB) with a cross-selection function, a fully automated sample injector, and a personal computer with a software for sample injection, data acquisition, and chemometric analysis.

Experimental Procedures.
Pieces of each sample (10 g) were placed in a beaker, soaked with 200 mL of pure water for 30 min, and then decocted for 30 min. e solution was filtered immediately. e residue was processed according to the abovementioned method twice. Afterward, all filtrates were combined. e obtained solution was placed into the special beaker of the electronic tongue and detected at room temperature.
Each sensor collected data from each sample for 120 s and was cleaned for 10 s. en, data were recorded by the data acquisition system. All assays were carried out in triplicate.

Pattern Recognition.
In this paper, PCA and LDA were used to differentiate C. songaricum originating from different places. PCA is a multivariate statistical method that reduces the dimensionality of data while retaining most of the variation in the data [21]. is approach was created before World War II but became widely used during the "Quantitative Revolution" in the 1960s [22]. New linear combinations of variables are created to accomplish the reduction. e combinations, called principal components, characterize the objects studied and satisfy certain statistical and mathematical conditions. us, samples can be displayed by few variables, and assessment of similarities and differences among samples is simplified. us, PCA is a suitable method to discriminate different samples and is extensively applied in food and drug analysis.
LDA is another commonly used technique for data discrimination and dimensionality reduction. LDA is strongly linked to regression analysis and analysis of variance (ANOVA), which also aims at expressing one dependent variable as a combination of other measurements or features [23]. e method maximizes the ratio of the between-class distance to the within-class distance to guarantee maximum discrimination. LDA has been used in numerous applications, such as image retrieval, microarray data classification, face recognition, and food and beverage discrimination [24][25][26][27]. Figure 2 shows the radar map of samples from different places in China. e sensor response of the electronic tongue varied with the change in geographical origins. Evidently, sensors BB, CA, and ZZ show strong signals to the samples. In particular, sensor BB exhibits the strongest response to the samples. Figure 2(a) shows the distinction among all samples from different places clearly. Signals from different samples show a considerable difference. Figure 2(b) illustrates the signals of every sample separately. Samples KX, TX, and HX are different from others, with the signals of sensor ZZ of these samples not exceeding 1000. Shapes of the radar maps of other samples are similar.

Principal Component
Analysis. e first discrimination model was established using PCA to visualize the different C. songaricum groups where possible. e accumulated explained variance was 89.6%, which was distributed in 79.5% (PC1) and 10.1% (PC2). Figure 3 shows the results of PCA score plot, and several trends are observed. Eight types of C. songaricum samples can be classified in general. Moreover, C. songaricum samples from Xinjiang are discriminated clearly between samples from other provinces. Similar samples appear in the same location of the graph. us, C. songaricum samples from Gansu, Ningxia, Qinghai, and Inner Mongolia are similar. e chemical constituents of these samples may be similar as well. Figure 4 shows the dispersion of C. songaricum samples by the LDA model. Compared with the PCA model, the LDA model shows a clearer discrimination among the eight types of C. songaricum. e explained variances by each discriminant function (DF) were 96.0% (DF1) and 2.9% (DF2). Each group of C. songaricum samples can be distinctly classified with others. As a result, the LDA model is a superior method to discriminate C. songaricum from different growing areas in China.

Linear Discriminant Analysis.
Given the good discrimination feature of the model, we used the LDA model to predict the geographical origin of unknown C. songaricum samples. As shown in Table 2, the prediction model can classify C. songaricum by geographical origins with 100% degree of accuracy.

Conclusions
We applied an electronic tongue to classify and predict C. songaricum samples from different places of origin. PCA and LDA were used for discrimination. e LDA model shows a clearer discrimination than the PCA model. e eight  groups of C. songaricum samples can be classified precisely using the LDA model. On this basis, we used the LDA model to predict the geographical origin of several unknown C. songaricum samples. A prediction model with 100% degree of accuracy was achieved. We present for the first time a method for the discrimination and geographical origin prediction of C. songaricum from different growing areas in China according to their flavor by an electronic tongue. e operations of data acquisition and processing are simpler and more convenient than the traditional chemical methods. e acquired information can be used for evaluating the quality of C. songaricum growing in China according to our previous work. Moreover, the identification and quality analysis method presented by us will be helpful for further study of C. songaricum. Further efforts should be focused on investigating the connection between the flavor and chemical constituents of C. songaricum samples and correlating electronic tongue signals with human perceptions of taste.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.   Journal of Analytical Methods in Chemistry 5