The present study investigated the maneuverability and reasonability of sensory analysis, which has been applied in TCM identification for a long time. Ten assessors were trained and generated the human panel to carry out the organoleptic evaluation of twenty-five batches of Sha-Ren samples. Accordingly, samples were scored from 0 (lowest) to 10 (highest) for sensory attributes. Based on this, samples were divided into three classes: high class (Yang-Chun-Sha from Guang-Dong), moderate class (Yang-Chun-Sha samples from Yun-Nan and Guang-Xi), and low class (Lv-Qiao-Sha from marketplaces). For further background, three instrumental approaches were employed: morphological measurement with three indices (longitudinal diameter, transverse diameter, and 100-fruit weight), GC for determination of bornyl acetate contents, and E-nose for aromatic fingerprint. It is demonstrated in the results that GC and E-nose analyses were in great agreement with organoleptic evaluation. It gives insights into further studies on searching better morphological indicators and improving discriminant model of E-nose.
Amomi Fructus, called “Sha-Ren” in Chinese, has been one of the commonly used herbs in Traditional Chinese Medicine (TCM) for more than 1,300 years, mainly for the treatment of gastrointestinal diseases. Besides, its widespread application in food industry also gains much attentions, not only in China, but in other Southeast Asian countries, including Thailand, Vietnam, Burma, Indonesia, and so forth. Its yearly consumption reaches more than 3.1 million kg and maintains a constant increasing trend, which reflects strong market demands and potential commercial opportunities. However, due to confusable species of Sha-Ren from different producing areas, its price varies remarkably more than a hundredfold. Therefore, it is urgent and necessary to develop a rapid and reliable approach to distinguish Sha-Ren from different classes. On the other hand, the adulteration also needs cautiousness, especially in clinical use, because incorrect or fake herbal medicines could result in low clinic effect or even poisoning. Many qualitative and quantitative methods have been employed in quality assessment of Sha-Ren from different species and habitats, for instance, GC-MS and IR analysis [
Macroscopic identification, as one of the five typical methods for TCM authentication [
The unique advantage of organoleptic analysis benefits from its capability of offering information perceived via human senses stimulating by a complex set of chemical compounds. For instance, odor is mainly perceived by the interactions of the volatiles with the olfactory epithelium in our nasal cavity [
In Pharmacopoeia of People’s Republic of China [
In the view of pros and cons of each technique, there are no universal and perfect methods for the simultaneous analysis of every volatile compound and the particular odor, which is essential for the development of distinguishing Sha-Ren rapidly and reliably. To the best of our literature survey, no study on organoleptic analysis of Sha-Ren coupled with GC and E-nose has been reported to date. Therefore, this study herein aims
Bornyl acetate standard (batch number 110759-200604) was purchased from National Institutes for Food and Drug Control (Beijing, China). Absolute ethyl alcohol and other reagents used were all at chromatographic grade and from Dikma Technologies Co., Ltd. (Beijing, China).
A total of 25 batches of Sha-Ren were collected: 19 batches of them were Yang-Chun-Sha from three habitats (Guang-Dong, Yun-Nan and Guang-Xi) and the rest 6 batches of them were Lv-Qiao-Sha from marketplaces. As mentioned above, the third species Hai-Nan-Sha were not obtained. All the samples have been authenticated via Professor Yong-Hong Yan from Department of Chinese Materia Medica of Beijing University of Chinese Medicine. They were packed in sealable plastic bags separately and stored at 4°C until analysis.
Based on observing, touching, and sniffing senses, organoleptic analysis was carried out by a human panel with ten assessors [
Three objective indices were chosen to represent the morphological characteristics: longitudinal diameter (LD), transverse diameter (TD), and 100-fruit weight (100 FW). According to published reports and experienced experts, these three indices summarized the size of Sha-Ren and expressed the extent of oiliness, which are closely related to the appearance and texture of Sha-Ren. What is more, these three could be measured into figures, which helped us conduct further analysis.
Method of coning and quartering were employed to take 20 Sha-Ren fruits. Then LD and TD values were measured by vernier caliper (Measuring Instrument & Cutting Tool Factory, Beijing, China) and afterwards the mean was calculated.
Also using method of coning and quartering but to take 100 Sha-Ren fruits, then 100 FW value was measured by BS-124S electronic analytical balance (Sartorius, Germany) in triplicate and the means was calculated as well.
All values should not be utilized until RSD < 3%.
According to previous reports, bornyl acetate is one of the main active volatile compounds [
Bornyl acetate standard was weighed precisely and dissolved in absolute ethyl alcohol to 11.90 mg/mL.
Around 1 g of pulverized Sha-Ren samples was added to a conical flask with 25 mL absolute ethyl alcohol and extracted with ultrasonic at 40 kHz for 30 min. After cooling down to the room temperature and compensating weight, the processed samples were filtered and stored at 4°C until analysis.
The GC system included an Agilent 7890A instrument (California, America) with a flame ionization detector (FID): GC conditions: Agilent DB-1 column, (0.25
Sha-Ren samples were firstly ground and sifted through a mesh of 850
The volatile compounds in the upper air from the sealed glass interacted with the MOG sensors and resulted in changes of their resistance, positively or negatively, which were recorded in the computer. The relative change in their resistance (
Friedman test, one of nonparametric tests, was introduced to analyze the 25 related samples and compare the consistency of the organoleptic evaluation by 20 panelists. Then logistics regression analysis with ordinal variables was employed to find out the differences of sensory characteristics among the samples from different species and habitats.
For data processing of morphological measurement, hierarchical cluster analysis (HCA) and discriminant analysis (DA) were both utilized to see how many groups those all 25 batches samples were divided into and whether it was in the agreement with organoleptic evaluation.
In the past decades, E-nose has been used in many fields and well-known as a promising approach for fast and noninvasive detection [
Simple correlation was applied to figure out the relationship among the experimental results of sensory and instrumental analyses, namely, organoleptic evaluation, GC, and E-nose. Afterwards, scatterplots and line charts were generated to see if there is any linear correlation. In the end, principal components analysis (PCA) was used to classify the samples based on those three groups of data.
The statistical analyses were all performed applying SPSS version 22.0 (SPSS/IBM, Beijing, China; Licensing Certificate No. 20150408-LSBJ) and Weka software (free access at website:
Odor is one of the key elements for identification of TCM. Since the pros and cons of sensory and instrumental approaches, organoleptic evaluation by 20 trained panelists was firstly carried out and then morphological measurement, GC, and E-nose were used to offer further background for further verification, which means digital and objective information of volatiles compounds and aromatic fingerprint.
The descriptive statistics including samples’ number, mean value, and minimum/maximum were obtained via Friedman Test (Table
Descriptive statistics via Friedman test of 25 samples by 10 panelists.
| Mean | Std. deviation | Minimum | Maximum | |
---|---|---|---|---|---|
Sam. 1 | 10 | 9.50 | .527 | 9 | 10 |
Sam. 2 | 10 | 8.90 | .994 | 7 | 10 |
Sam. 3 | 10 | 9.40 | .699 | 8 | 10 |
Sam. 4 | 10 | 9.50 | .527 | 9 | 10 |
Sam. 5 | 10 | 8.10 | .738 | 7 | 9 |
Sam. 6 | 10 | 9.00 | .943 | 7 | 10 |
Sam. 7 | 10 | 8.80 | .919 | 8 | 10 |
Sam. 8 | 10 | 9.10 | .994 | 7 | 10 |
Sam. 9 | 10 | 5.40 | .843 | 4 | 7 |
Sam. 10 | 10 | 5.10 | .738 | 4 | 6 |
Sam. 11 | 10 | 4.30 | .675 | 3 | 5 |
Sam. 12 | 10 | 5.10 | .876 | 4 | 6 |
Sam. 13 | 10 | 4.50 | .972 | 3 | 6 |
Sam. 14 | 10 | 5.00 | 1.054 | 3 | 6 |
Sam. 15 | 10 | 4.80 | .919 | 3 | 6 |
Sam. 16 | 10 | 3.80 | .919 | 2 | 5 |
Sam. 17 | 10 | 2.30 | 1.252 | 0 | 4 |
Sam. 18 | 10 | 2.50 | .850 | 1 | 4 |
Sam. 19 | 10 | 2.80 | 1.135 | 0 | 4 |
Sam. 20 | 10 | 1.60 | .843 | 0 | 3 |
Sam. 21 | 10 | 1.50 | .707 | 1 | 3 |
Sam. 22 | 10 | 2.50 | .707 | 1 | 3 |
Sam. 23 | 10 | 1.30 | .949 | 0 | 3 |
Sam. 24 | 10 | 1.50 | .972 | 0 | 3 |
Sam. 25 | 10 | 1.20 | .919 | 0 | 3 |
According to logistic regression analysis with ordinal variables, the probability of each kind of samples could be forecasted, which is how many scores they will earn through this human panel. For instance, the total number of Yang-Chun-Sha from Guang-Dong province is 80 (5 + 15 + 32 + 28) as shown in Table
Cell information of logistic regression analysis with ordinal variables from 25 samples by 10 panelists.
Frequency | Score | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Habitats | Species | |||||||||||
0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | ||
GuangDong | YangChunSha | |||||||||||
Observed | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 5 | 15 | 32 | 28 | |
Expected | .000 | .000 | .000 | .000 | .000 | .000 | .000 | 5.000 | 15.000 | 32.000 | 28.000 | |
Pearson Residual | .000 | .000 | .000 | .000 | .000 | .000 | .000 | .000 | .000 | .000 | .000 | |
| ||||||||||||
YunNan | YangChunSha | |||||||||||
Observed | 0 | 0 | 0 | 2 | 16 | 19 | 12 | 1 | 0 | 0 | 0 | |
Expected | .055 | .247 | 1.149 | 4.479 | 11.597 | 17.566 | 14.056 | .850 | .000 | .000 | .000 | |
Pearson Residual | −.235 | −.499 | −1.085 | −1.228 | 1.475 | .425 | −.647 | .164 | .000 | .000 | .000 | |
| ||||||||||||
GuangXi | YangChunSha | |||||||||||
Observed | 2 | 2 | 10 | 16 | 14 | 10 | 6 | 0 | 0 | 0 | 0 | |
Expected | .471 | 2.036 | 8.077 | 18.863 | 18.222 | 8.972 | 3.215 | .145 | .000 | .000 | .000 | |
Pearson Residual | 2.235 | −.025 | .727 | −.796 | −1.185 | .372 | 1.597 | −.381 | .000 | .000 | .000 | |
| ||||||||||||
Market | LvQiaoSha | |||||||||||
Observed | 7 | 21 | 21 | 11 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Expected | 8.280 | 19.830 | 20.631 | 8.329 | 2.171 | .583 | .168 | .007 | .000 | .000 | .000 | |
Pearson Residual | −.479 | .321 | .100 | .997 | −1.501 | −.767 | −.411 | −.085 | .000 | .000 | .000 |
Link function: Logit.
To summarize the results of organoleptic evaluation by 10 panelists, all 25 batches of Sha-Ren samples were divided into three groups, namely, high class (Yang-Chun-Sha from Guang-Dong), moderate class (Yang-Chun-Sha samples from Yun-Nan and Guang-Xi), and low class (Lv-Qiao-Sha from marketplaces). It resembles the concept of TCM experts that Yang-Chun-Sha from Guang-Dong is the geoauthentic medicine of better quality and clinic effects.
HCA applying furthest neighbor cluster method shows that all samples were classified into four groups (Figure
HCA vertical icicle diagram responses to samples based on morphological measurements with three indices (LD, TD, and 100 FW).
To the next step, DA was used to find out whether these three morphological indices could help to get a satisfying identification of Sha-Ren samples. From the test of functions, the values of Wilks’ Lambda are 0.295, 0.664, and 0.989, respectively. Sig. values of them are all higher than 0.01. Therefore, this discriminant analysis could not be able to separate the samples efficaciously, which stays on the same side of HCA results.
Bornyl acetate in extracted samples was identified by comparison of their retention time with that obtained from the chromatograms of standard substance. The content of bornyl acetate in each sample was calculated using standard curve of bornyl acetate and illustrated in Table
Experimental results of Sha-Ren samples.
Batch number | Sample name | Species | Habitats | Score of human panel | Bornyl acetate content | S6-sensor response |
---|---|---|---|---|---|---|
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 9.5 | 3.82 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 8.9 | 6.64 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 9.4 | 5.68 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 9.5 | 1.74 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 8.1 | 5.03 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 9.0 | 1.52 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 8.8 | 3.66 | 0.04 |
| GD-YCS | Yang-Chun-Sha | Guang-Dong | 9.1 | 1.35 | 0.04 |
| YN-YCS | Yang-Chun-Sha | Yun-Nan | 5.4 | 0.88 | 0.03 |
| YN-YCS | Yang-Chun-Sha | Yun-Nan | 5.1 | 1.06 | 0.03 |
| YN-YCS | Yang-Chun-Sha | Yun-Nan | 4.3 | 0.48 | 0.03 |
| YN-YCS | Yang-Chun-Sha | Yun-Nan | 5.1 | 3.71 | 0.03 |
| YN-YCS | Yang-Chun-Sha | Yun-Nan | 4.5 | 4.04 | 0.03 |
| GX-YCS | Yang-Chun-Sha | Guang-Xi | 5.0 | 1.24 | 0.03 |
| GX-YCS | Yang-Chun-Sha | Guang-Xi | 4.8 | 4.31 | 0.03 |
| GX-YCS | Yang-Chun-Sha | Guang-Xi | 3.8 | 0.80 | 0.03 |
| GX-YCS | Yang-Chun-Sha | Guang-Xi | 2.3 | 0.73 | 0.03 |
| GX-YCS | Yang-Chun-Sha | Guang-Xi | 2.5 | 0.71 | 0.04 |
| GX-YCS | Yang-Chun-Sha | Guang-Xi | 2.8 | 0.69 | 0.03 |
| M-LQS | Lv-Qiao-Sha | Marketplaces | 1.6 | 0.68 | 0.03 |
| M-LQS | Lv-Qiao-Sha | Marketplaces | 1.5 | 0.70 | 0.03 |
| M-LQS | Lv-Qiao-Sha | Marketplaces | 2.5 | 0.74 | 0.03 |
| M-LQS | Lv-Qiao-Sha | Marketplaces | 1.3 | 0.72 | 0.03 |
| M-LQS | Lv-Qiao-Sha | Marketplaces | 1.5 | 0.52 | 0.03 |
| M-LQS | Lv-Qiao-Sha | Marketplaces | 1.2 | 0.64 | 0.03 |
In the estimation of discriminant models using three different classifiers (NBN, RBF, and RF), two validation methods were applied and distinguishing positive rate was calculated, which were tenfold cross-validation in the training set and external test set validation in the test set.
Through BA feature extraction method performed in Weka software, six MOG sensors, namely, S2, S3, S6, S8, S10, and S12, were selected to generate the optimum data set. However, based on this discriminant model, only Yang-Chun-Sha from Guang-Dong could be distinguished from the other samples successfully, which leads to two categories based on E-nose responses. It is in similar situation of organoleptic evaluation results, where it sheds light on the concept again that Yang-Chun-Sha is the geoauthentic medicine.
As shown in Table
Distinguishing positive rates of three classifiers (NBN, RBF, and RF) of original and optimum data set.
Cliassifer | Original data set | Optimum data set | ||
---|---|---|---|---|
Tenfold cross-validation | External test set validation | Tenfold cross validation | External test set validation | |
NBN | 54 | 56 | 78 | 84 |
RBF | 66 | 72 | 78 | 84 |
RF | 64 | 76 | 78 | 84 |
For the simple correlation analysis, the organoleptic analysis showed positive correlation with GC and E-nose, with correlation index of 0.650 and 0.807, respectively. However, according to the scatterplots and line graphs, no obvious linear regression was found. But it is easy to see that the curves of organoleptic analysis and GC are in the same trend.
As for PCA, samples were placed in three zones (Figure
PCA score plot responses to Sha-Ren samples with PC1 and PC2.
Twenty-five batches of Sha-Ren samples were collected from Guang-Dong, Guang-Xi, and Yun-Nan and marketplaces, which are Yang-Chun-Sha (
The authors report no conflicts of interest related to this manuscript.
Dong Xu and Yan Lin contributed equally to this project.
This project was financially supported by the National Natural Science Foundation of China (Grants nos. 81403054, 81573542, and 81202682).