HPLC fingerprint analysis combined with chemometrics for pattern recognition of ginger

Abstract

Context: Ginger, the fresh rhizome of Zingiber officinale Rosc. (Zingiberaceae), has been used worldwide; however, for a long time, there has been no standard approbated internationally for its quality control.

Objective: To establish an efficacious and combinational method and pattern recognition technique for quality control of ginger.

Methods: A simple, accurate and reliable method based on high-performance liquid chromatography with photodiode array (HPLC-PDA) detection was developed for establishing the chemical fingerprints of 10 batches of ginger from different markets in China. The method was validated in terms of precision, reproducibility and stability; and the relative standard deviations were all less than 1.57%. On the basis of this method, the fingerprints of 10 batches of ginger samples were obtained, which showed 16 common peaks. Coupled with similarity evaluation software, the similarities between each fingerprint of the sample and the simulative mean chromatogram were in the range of 0.998–1.000. Then, the chemometric techniques, including similarity analysis, hierarchical clustering analysis and principal component analysis were applied to classify the ginger samples.

Results and conclusion: Consistent results were obtained to show that ginger samples could be successfully classified into two groups. This study revealed that HPLC-PDA method was simple, sensitive and reliable for fingerprint analysis, and moreover, for pattern recognition and quality control of ginger.

• Chemical fingerprint
• chemometrics
• ginger
• HPLC-PDA
• pattern recognition

Introduction

Zingiberis Rhizoma Recens, Shengjiang in Chinese, is the fresh rhizome of Zingiber officinale Rosc. (Zingiberaceae) (China Pharmacopoeia Committee, 2010) and is well known as both a medicinal and edible plant. It has been used as a herb in traditional Chinese medicine (TCM) and food for thousands of years. As a TCM, ginger could induce perspiration, dispel cold, warm the stomach and arrest vomiting, resolve phlegm and relieve cough, and resolve fish or crab poisoning. Moreover, as an edible plant, it could flavor food, stimulate digestion and promote the functional activity of the stomach. Therefore, ginger has been widely used all over the world. However, for a long time, there has been no standard approbated internationally for the quality control of ginger, which has seriously affected the development and exchange of ginger and its related medicinal and edible products. In addition, to date, few investigations have described the quality assessment of ginger in detail.

According to Chinese medicine theory, the whole components in crude herbs are responsible for their beneficial medicinal effects, which make the quality control of herbal products very difficult. Traditionally, one or two markers or active components in herbs or herbal mixtures are used to assess the authenticity and quality of the complex TCMs. This strategy has been proved to be insufficient for the quality control of TCMs and their preparations because it does not evaluate all chemical components present in the chromatographic profile. With the development of analytical technique, chromatographic fingerprints have been widely used for the authentication and quality control of TCMs (Gong et al., 2003; Liang et al., 2004). Chromatography fingerprint techniques can be used to characterize both the marker compounds and the unknown components in a complex system, a strategy recommended to assess the quality and consistency of botanical products by the US Food and Drug Administration, the European Medicines Evaluation Agency, and State Food and Drug Administration of China (SFDA) (The European Agency for the Evaluation of Medicinal Products, 2006; State FDA, 2000; US FDA, 2004). In the past several years, chromatographic fingerprints have been established using high-performance liquid chromatography (HPLC), HPTLC, GC and CE, and they have been recognized as rapid, reliable methods for the identification and recognition of herbal medicines (Duan et al., 2012; Kong et al., 2009a; Wagner & Bladt, 2001; Wagner et al., 2009). Of these methods mentioned above, HPLC is the most popular and is widely used for fingerprint analysis. Although visually distinguishing the differences between chromatograms is possible, the process is subjective and not quantitative. In addition, the fingerprint chromatograms consist of complex multivariate data sets due to the complexity of herbal medicines, so minor differences between very similar chromatograms might be missed (Xu et al., 2006). Thus, the chemometric methods, such as similarity analysis (SA), hierarchical clustering analysis (HCA) and principal component analysis (PCA), etc., should be taken into consideration for reasonable pattern recognition, further for quality control of these herbal medicines (Duan et al., 2012; Kong et al., 2008, 2009b).

In this study, a simple, sensitive and reliable HPLC photodiode array (PDA) method was developed for establishing the chemical fingerprints of 10 batches of ginger samples from various markets in China. Combined with chemometric methods, such as SA, HCA and PCA, these ginger samples were efficiently differentiated. The aim was to establish an efficacious method and pattern recognition technique for quality control of ginger, further for providing some insight and reference for quality control of other TCMs.

Materials and methods

Materials and reagents

Ten batches of ginger samples were collected from different vegetable markets or supermarkets in Beijing, China, during July and August 2012. They grew and were harvested in Shandong, China, and have been identified by Prof. Bengang Zhang, Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China. Some voucher specimens were preserved in the Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China.

Standard of 6-gingerol (purity >98%) was purchased from Express Technology Co. Ltd. (Beijing, China). Acetonitrile of HPLC grade was bought from Fisher Chemicals (Pittsburg, PA). Wahaha purified water was used throughout the study. Other chemicals were all of analytical grade obtained from the Beijing Chemical Factory (Beijing, China). All the solutions were filtered through a 0.22 μm membrane (Schleicher & Schuell, Dassel, Germany).

Apparatus and chromatographic conditions

Chromatographic analysis was performed on a Waters 2690 Alliance system (Waters, Milford, MA) equipped with a binary solvent delivery pump, an online vacuum degasser, an auto sampler, a thermostat column compartment and a PDA detector and connected to a Waters Empower 2 data station. The chromatographic separation was performed using a Rainbow Kromasil C₁₈ column (4.6 mm × 250 mm, 5 μm), operated at 25 °C. The mobile phase consisted of acetonitrile (solvent A) and 0.1% phosphate in water (solvent B) with a linear gradient procedure as listed in Table 1. The flow rate was 0.8 mL min⁻¹, and the sample injection volume was 20 μL. The ultraviolet (UV) absorbance was monitored at the range of 210–400 nm, and the detection wavelength was 280 nm. Each run took 110 min.

Table 1. Linear gradient elution procedure.

t/min	A/%	t/min	A/%
0	5	70	70
10	25	90	90
20	50	100	100
30	50	110	100
60	55

Preparation of standard solution

The standard solution was prepared by adding an accurately weighed amount of 6-gingerol standard into a volumetric flask and dissolving with 10 mL methanol. It was stored in a dark glass bottle at 4 °C and was stable for at least 1 month.

Preparation of sample solution

Each of the finely cut samples (50 g) of ginger was accurately weighed and added to 100 mL methanol. The mixture was precisely weighed and placed into an ultrasonic bath (40 kHz) for 30 min. After the mixture cooled to the room temperature, methanol was added to compensate for the lost weight during the extraction, and then the extracted solution was filtered. The filtrate was further filtered through a 0.22-μm membrane and injected into the HPLC system for analysis.

Validation of HPLC method

The HPLC method was validated in terms of precision, reproducibility and stability according to ICH guidelines (ICH Topic Q2B, 1996). The precision of HPLC-PDA method was calculated as the relative standard deviation (RSD) of six repeated runs. The reproducibility of the method was evaluated by running six replicate samples prepared independently in a single day. Sample stability was monitored by analyzing the same sample solutions for 1 d at an interval of 4 h. The value of RSD was taken as the appraise index.

Chemometric analysis for quality evaluation

Similarity analysis

The HPLC-PDA fingerprints of 10 batches of ginger samples was first evaluated by SA using professional software namely Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine (Version 2004A) composed by the Chinese Pharmacopoeia Committee, which was recommended by SFDA of China. The simulative mean chromatogram was generated by this software as a representative reference chromatogram for a group of chromatograms of these ginger samples from different markets. Then, the values of SA between each chromatogram of ginger sample and the reference chromatogram was calculated, respectively.

Hierarchical clustering analysis

In order to evaluate the resemblance and differences in these samples, HCA was performed on the common peak areas in the HPLC fingerprints. HCA is a multivariate analysis technique that can be used to divide samples into groups (Kannel et al., 2007). In this study, HCA of the ginger samples was performed using DataLab 2.7 software (Pressbaum, Austria) from http://www.lohninger.com/datalab/.

Principle component analysis

For further discriminating the investigated samples, PCA (Kannel et al., 2007; Karisa et al., 2005), a sophisticated technique widely used for reducing the dimensions of multivariate problems, was carried out based on the differences in the samples. This technique can reduce the dimensionality of the original data set by explaining the correlation amongst a large number of variables in terms of a smaller number of underlying factors (principal components or PCs) without losing much information (Jackson, 1991). In this study, PCA was performed on the common peak areas from the HPLC fingerprints using DataLab 2.7 software.

Results

Optimization of HPLC condition

To give the most chemical information and best separation in the chromatograms, the mobile phase and its flow rate, conditions for elution, column temperature and detection wavelength were investigated in this study.

The chromatographic conditions were optimized using both 6-gingerol standard and ginger samples. The different ratio of water to acetonitrile in the mobile phase could elute different components, so the linear gradient elution was selected. In this study, the addition of phosphate in the mobile phase water provided the best resolution and separation of the components. From the results of comparative study of column temperature of 20 °C, 25 °C and 30 °C and flow rate of 0.6, 0.8 and 1.0 mL min⁻¹, the flow rate was set at 0.8 mL min⁻¹ when the column temperature was kept at 25 °C.

Furthermore, the UV absorbance was monitored at the range of 210–400 nm; and by comparison, 280 nm was selected as the detection wavelength.

Under the optimal conditions, the components in the extracts of these ginger samples were well separated as shown in Figure 1.

Figure 1. HPLC-PDA chromatograms of (A) standard compound and (B) ginger sample.

HPLC method validation

Precision

Six successive injections of the same sample solution were processed using Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine for the evaluation of the precision of the apparatus. The RSDs (%) regarding to the peak area of 16 common peaks were 1.36, 0.44, 0.25, 0.18, 0.00, 0.34, 0.25, 0.33, 0.41, 0.48, 0.50, 0.52, 0.57, 0.54, 0.53 and 0.58, and the similarities between the reference HPLC fingerprint and the sample chromatogram were 1, 1, 1, 1, 1 and 1, indicating satisfactory precision of the apparatus.

Stability

Six injections of the same sample solution stored for 0, 4, 8, 12, 16 and 20 h were processed using Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine for the evaluation of sample stability. RSDs (%) regarding to the peak area of 16 common peaks were 1.57, 0.55, 0.27, 0.20, 0.00, 0.44, 0.17, 0.22, 0.34, 0.40, 0.43, 0.46, 0.52, 0.50, 0.57 and 0.61, and the similarities between the reference HPLC fingerprint and the sample chromatogram were 1, 1, 1, 1, 1 and 1, indicating that the sample solution was stable within the tested time period.

Reproducibility

Six solutions from the same sample were prepared using the same method and were injected individually to evaluate the repeatability of the method. The data was processed using Similarity Evaluation System for Chromatographic Fingerprint of Traditional Chinese Medicine. RSDs (%) regarding to the peak area of 16 common peaks were 0.73, 0.30, 0.16, 0.06, 0.00, 0.11, 0.15, 0.18, 0.22, 0.25, 0.26, 0.26, 0.28, 0.29, 0.44 and 0.39, and the similarities between the reference HPLC fingerprint and the sample chromatogram were 0.999, 0.998, 0.998, 0.999, 1.000 and 1.000, showing suitable repeatability of the analytical method.

HPLC fingerprints of ginger

To obtain the standard fingerprint, 10 batches of samples were analyzed. Chromatograms of these samples are shown in Figure 2(A), and the reference fingerprint of ginger is shown in Figure 2(B). The peaks that existed in all 10 samples with reasonable heights and good resolution were assigned as “characteristic peaks” for the identification of the plant. The 16 characteristic peaks that appeared within the elution time are shown in Figure 2(B). Peak 5 (6-gingerol), which was one of the most important active constituents of ginger, was chosen to calculate the relative retention time and relative peak area (RPA) of all the other peaks.

Figure 2. HPLC fingerprints of (A) 10 batches of ginger samples (S1–S10) and (B) the reference chromatogram.

Quality evaluation by SA, HCA and PCA

Similarity analysis

It was necessary that chromatographic fingerprint of ginger from various markets should be evaluated by their similarities, which came from the calculation on the correlative coefficient of original data. Thus, the correlation coefficients between each chromatogram of ginger samples and the simulative mean chromatogram, which was the median of all chromatograms, were shown in Table 2. All the correlation coefficients were larger than 0.918. The closer the cosine values were to 1, the more similar the two chromatograms were (Wei et al., 2010). If the similarity value over a certain value, 0.9 for example, were regarded as the threshold for qualification, it was easy to identify the qualified samples based on the chromatographic fingerprint. The result indicated that the samples shared different correlation coefficients of similarities, showing the internal quality of these samples were different.

Table 2. Raw herbs tested in this work and the similarity in their chromatography.

Sample number	Sources	Origins	Collected date	Similarity
S1	MerryMart supermarket	Zingiber officinale Rose.	7 January 2012	0.987
S2	Lanyuan vegetable store	Zingiber officinale Rose.	7 January 2012	0.998
S3	Xinglong free market	Zingiber officinale Rose.	7 January 2012	0.918
S4	MerryMart supermarket	Zingiber officinale Rose.	8 January 2012	0.983
S5	Xinglong vegetable store	Zingiber officinale Rose.	8 January 2012	0.982
S6	MerryMart supermarket	Zingiber officinale Rose.	9 January 2012	0.989
S7	Hualian supermarket	Zingiber officinale Rose.	9 January 2012	0.971
S8	Chaoshifa supermarket	Zingiber officinale Rose.	9 January 2012	0.995
S9	Baijiali community store	Zingiber officinale Rose.	9 January 2012	0.983
S10	Xinglong vegetable store	Zingiber officinale Rose.	9 January 2012	0.992

Hierarchical cluster analysis

In order to assess this resemblance and differences of these samples, a hierarchical agglomerative clustering analysis of ginger samples was performed based on the RPAs of all the 16 common chromatographic peaks. The RPAs of the common constituents in 10 batches of ginger samples from various markets formed a matrix of 16 × 10. The results of HCA are shown in Figure 3, from which the quality characteristics were revealed more clearly. Supposing an appropriate distance level was chosen, the samples could be classified into two quality clusters. Except for S2, cluster I was formed by the samples collected from supermarkets. The cluster II consisted of the samples collected from free markets or vegetable stores, except S4 and S6. These results showed that the quality characteristics of the samples from the supermarkets were more similar, while, the quality characteristics of the samples from the free markets or vegetable stores were more similar. S2 was classified as cluster I indicated that its quality characteristic was more similar to those from supermarkets rather than to those from free markets or vegetable stores. S4 and S6 were classified as cluster II, which suggested that their quality were more similar to those purchased from free market or vegetable stores. This was probably because of some conceivable reasons, such as the differences of batches and harvest time or that they had been stored too long in the warehouse of the supermarkets. From Figure 3 and Table 2, it could be seen that when the similarities of two samples were closer, while the distance between them in Figure 3 was smaller, and vice versa. It indicated that the results of HCA were consistent with those of SA.

Figure 3. Dendrogram of clustering of 10 batches of ginger samples (S1–S10).

Principle component analysis

In order to amplify the difference in the samples and easily discriminate the samples, PCA, a method for feature extraction and dimensionality reduction, was carried out (Chen et al., 2008). The PCA computation was implemented by performing singular value decomposition on the data array of the fingerprints, which consisted of a 10 × 16 data matrix, with each row representing a plant sample and each column containing the values of 16 characteristic peak areas. To display the points on two PCs, PC1 and PC2 (the first and second PC) were chosen to represent the information. As shown in Figure 4, PCA displayed the result that 10 batches of samples were classified into two groups, similar to the results of HCA. Moreover, the results from PCA were largely consistent with that of the similarity evaluation. Group I consisted of samples purchased from supermarkets and S2. It suggested that they were associated with similar internal quality. Group II consisted of the samples purchased from free markets or vegetable stores and S4 and S6. It could be indicated that they had similar internal quality. These results were corresponding with the SA and same to HCA. The results of HCA and PCA could be validated each other and provided more references for the quality evaluation of ginger.

Figure 4. Scores plot of PCA for different ginger samples on the first two PCs with the original peak areas of the 16 common compounds as input data.

Discussion and conclusion

A simple, reliable and accurate HPLC-PDA method has been developed for the fingerprint analysis of ginger. The fingerprint of the ginger sample showing 16 common peaks represented the characteristics of the herb’s constituents and provided an accurate method for the quality control of ginger. It was difficult in classifying the ginger samples according to different sources. Chemometric methods should be applied with the HPLC fingerprint techniques for the discrimination of the ginger samples.

This work has shown that chemometric techniques such as SA, HCA and PCA were able to classify samples objectively and successfully. The results suggested that 10 batches of ginger samples could be classified into two groups. One consisted of samples purchased from supermarkets, the other consisted of samples purchased from free markets or vegetable stores, which could to some extent reflect the quality differences of theses samples. Meanwhile, the cosine and correlation coefficient values of 10 batches of samples were more than 0.9. If 0.9 is set as an appropriate threshold, it is easy to identify ginger based on the chromatographic fingerprint. The method established in this study was simple, sensitive and reliable and could be used for the evaluation and quality control of ginger and its related products. It also provided an important reference for the establishment of the method for pattern recognition and quality control of other TCMs.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article. The authors are grateful for the support from the National Science Foundation of China (81274072) and Specialized Research Fund for the Doctoral Program of Higher Education (20121106120029).