Analysis of Quality Differences in Radix Dipsaci before and after Processing with Salt Based on Quantitative Control of HPLC Multi-Indicator Components Combined with Chemometrics

Radix Dipsaci (RD) is the dry root of the Dipsacus asper Wall. ex DC., which is commonly used for tonifying the kidney and strengthening bone. The purpose of this study was to analyze the difference between raw and salt-processed RD from the chemical composition comprehensively. The fingerprints of raw and salt-processed RD were established by HPLC-DAD to determine the contents of loganin (LN), asperosaponin VI (AVI), caffeic acid (CaA), dipsanoside A (DA), dipsanoside B (DB), chlorogenic acid (CA), loganic acid (LA), isochlorogenic acid A (IA), isochlorogenic acid B (IB), and isochlorogenic acid C (IC). The results showed that after processing with salt, the components with increased contents were LA, CaA, DA, and AVI, and the components with decreased contents were CA, LN, IB, IA, IC, and DB. Then, the chemometric methods such as principal component analysis (PCA) and fisher discriminant analysis (FDA) were used to evaluate the quality of raw and salt-processed RD. In the classification of raw and salt-processed RD, the order of importance of each chemical component was LA > DB > IA > IC > IB > LN > CA > DA > AVI > CaA. These integrated methods successfully assessed the quality of raw and salt-processed RD, which will provide guidance for the development of RD as a clinical medication.

Chemometrics is an emerging interdisciplinary discipline formed by the combination of mathematics, statistics, computer science, and chemistry and is an important means of material-based research of traditional Chinese medicine (TCM).It can introduce multivariate analytical methods into chemical research, process and analyze chemical measurement data in multiple ways, create and optimize various chemical models, and extract the components, structures, and other related information of related substance systems from complex chemical measurement data to the maximum extent.At present, the combination of spectroscopic data and chemometric methods has been used by many scholars to research TCM [11][12][13].In this study, HPLC-DAD fngerprints of RD were established, and 10 diferent components in RD before and after processing with salt were selected for determination and combined with the chemometric methods such as principal component analysis (PCA) and fsher discriminant analysis (FDA) to establish a more comprehensive and quantitative chemical pattern identifcation and quality evaluation method for the samples of RRD and SRD, providing a certain scientifc basis for the later development of studies on spectrum-efect relationship.

Sample Collection.
RRD was purchased from diferent origins in China and was authenticated by Professor Weihong Ge (School of Pharmacy, Zhejiang Chinese Medical University).SRD was prepared from RRD. Te processing method of SRD was as follows: mixed RRD with salt water evenly, made it moist, stir-fried for 10 minutes at 160 °C in a frying pan, took it out, and let it cool (every 100 kg RRD with 2 kg salt).Te samples were stored in the herbarium of Zhejiang Chinese Medical University Chinese Medicine Yinpian Co., Ltd.(Hangzhou, China).Te sample information is shown in Table 1.

Methods
Te HPLC method for RD was formulated and optimized by our group in a previous work and has been published in other articles by Wu et al. [14].In this paper, we followed this method for the determination of the content of each component in RD.Te sample powder (80 mesh) was weighed 0.5 g accurately and placed in a conical fask with a stopper, and 25 mL of 80% methanol was added and weighed accurately.Te solution was ultrasonicated (power 300 W and frequency 50 kHz) for 30 min and then weighed again.80% methanol was used to make up the lost weight, and the sample solution was obtained by fltering through a 0.45 μm microporous membrane.

HPLC Methodological Investigation.
Linear relationship investigation took the standard solution and diluted it with 80% methanol to make 10 gradient concentration solutions from high to low.According to the chromatographic conditions under "3.2," injected 10 μL of the diferent concentration standard solution.Drew a standard curve with peak area (y) and concentration (x, μg/mL) to calculate the regression equation.At the same time, we determined the limit of detection (LOD) of the injection concentration when S/N = 3 and the limit of quantifcation (LOQ) of the injected concentration when S/N = 10.
Precision took the same R1 sample solution and continuously injected for 6 times according to "3.2"; repeatability took the R1, prepared 6 samples of the sample solution in parallel, and injected according to "3.2"; stable properties took the R1 sample solution and injected according to "3.2" at 0, 2, 4, 8, 12, 16, and 24 hours, respectively (sample solutions were stored at room temperature).Te peak areas of the three investigation items were recorded, respectively, and the relative retention time (RRT) International Journal of Analytical Chemistry and relative peak area (RPA) RSD of each common peak were calculated with AVI as the reference peak.
Te recovery rate of sample addition weighed 6 RD samples powders (R1, S1) with known content, respectively, and added 10 standard substances, respectively (according to the content of the component in the sample, we added a certain volume of corresponding dilutions and then evaporated the solvent).We prepared the tested solution according to "3.1," injected according to "3.2," and analyzed it to obtain the average value and RSD of the recovery rates of 10 components.

Establishment of HPLC Fingerprints and Content
Determination.We took each batch of RRD and SRD to prepare the sample solution according to "3.1," injected according to "3.2," then imported the collected chromatographic data into "Evaluation of Similarity of Chinese Medicine Fingerprints Software," respectively, generated the contrast map of raw and salt-processed RD, calculated the similarity, pipetted 10 μL of the standard solution to inject it into HPLC, and used the standard data to determine the content of LA, CA, LN, AVI, CaA, IA, IB, IC, DA, and DB in each sample.

Chemometric Analysis.
Te data of 10 components in each sample obtained by the content determination were imported into SPSS for PCA and FDA, and the diference in quality between the RRD and SRD was analyzed.

Results of Methodological Investigation.
Te developed method was used to evaluate the linear range, recovery rate, precision, repeatability, and stability of the method for the determination of 10 components.Table 2 shows that the r of 10 components was all greater than 0.999 in the linear range, presenting a good linear relationship, which meets the experimental requirements.Te results of precision, repeatability, and stability showed that the RSD of RRT and RPA were both less than 3%, and the similarity was both greater than 0.995, indicating that the method could be used for HPLC detection of RD.Te sample recovery rate results in Table 3 showed that the recovery rates of the 10 components in the raw and salt-processed RD were all within the range of 95% to 100%, indicating that the accuracy was good and met the experimental requirements.

HPLC Fingerprints of Samples.
Te fngerprints of the obtained RRD and SRD are shown in Figures 1 and 2. Te results in Tables 4 and 5 showed that the similarity of the fngerprints of raw and salt-processed RD was above 0.900, respectively, and a total of 25 peaks were obtained.Figure 3 shows that compared with the results of the reference solution, 10 components were identifed, namely, peak 6-LA; peak 8-CA; peak 9-CaA; peak 10-LN; peak 11-IB; peak 12-IA; peak 14-IC; peak 16-DB; peak 17-DA; peak 20-AVI.
After processing, the contents of 10 components in the corresponding batches of raw and salted-processed RD were signifcantly diferent.As shown in Figure 4, the contents of LA, CaA, DA, and AVI in SRD were higher than RRD, the average change rates were 28.80%, 74.78%, 21.38%, and 18.99%, respectively, and the contents of CA, LN, IB, IA, IC, and DB in SRD were lower than those in RRD, with an average change rate of −23.28%, −30.96%, −36.30%, −21.49%, −29.49%, and −45.37%, respectively.Tis was due to the change in content caused by processing with salt, and it was speculated that the content of phenolic acids and iridoid glycosides in RD might be reduced due to the conversion and degradation of the components after heating.International Journal of Analytical Chemistry   8, the principal components were extracted with the eigenvalue λ > 1, λ 1 � 4.324, the contribution rate was 43.237%, λ 2 � 1.512, the contribution rate was 15.122%, λ 3 � 1.395, the contribution rate was 13.950%, and the contribution rate of the frst principal component was the largest, indicating that it contains the most information.
When the number of principal components was 3, the cumulative contribution rate reached 72.310%, and it indicated that the frst three principal components could represent most of the information data about the diference between the raw and salt-processed RD.Te frst, second, and third principal components were used as the coordinate system, and the three-dimensional map of each compound was obtained by projection.As shown in Figure 5, the 10 compounds were divided into 2 categories, one of which was LA, CaA, DA, and AVI and the other type was CA, LN, IB, IA, IC, and DB.Tis result was consistent with the change law of the content before and after processing with salt, the former was the components whose content increased after processing with salt, and the latter was the components whose content decreased.Te component loading matrix can explain the contribution rate of each variable to the principal component.Te greater the absolute value of the compound loading, the greater the contribution to the principal component, indicating that it is more important in the quality control of decoction pieces.According to the data in Te above discriminant functions were used to backsubstitute the classifcation.As shown in Figure 6, the samples of raw and salt-processed RD could be well differentiated in the discriminant analysis scatterplot.At the same time, the discriminants of RRD and SRD were consistent with the actual, and the accuracy rates were both 100%.

Discussion
Te HPLC fngerprints of RD before and after processing with salt were established.Tere were 25 common peaks in the fngerprints, and the contents of 10 components were determined.Te results showed that the contents of LA, CaA, DA, and AVI increased, while the contents of CA, LN, IB, IA, IC, and DB decreased after processing with salt.No new or disappeared components were found in the RD before and after processing with salt, and it was speculated that there might have been intercomponent transformations.For example, the increase in LA content may have been caused by the addition of the -COOH group to LN, which was also consistent with the decrease of LN and organic acids (CA, IB, IA, and IC).Te conversion between DA and DB led to an increase in DA content.Te study of the transformation between these components was also an important part of the processing mechanism of TCM.Whether the change in composition after processing with salt was caused by heating or salt processing during frying has not been systematically studied in this section.
Subsequent systematic studies will be carried out on the efect of excipient salt on the composition.
Chemometrics, also known as chemical statistics, is a branch of chemistry that combines mathematics, statistics, computer science, and chemistry, which the most important feature of chemistry is the introduction of multivariate analysis methods into chemical research and the multivariate processing and analysis of chemical measurement data.Chemometrics include measurement tests, chemical pattern recognition, regression analysis, and multivariate correction [15,16].PCA is an unsupervised pattern recognition analysis which uses the idea of reducing the dimension of the data matrix to convert the original indicators into several comprehensive indicators through a linear transformation under the premise of losing a small amount of information, so as to simplify datasets and visualize diferences between samples [17].Since there is no human involvement in the analysis process, and the calculation model is based on the state of the original variables, PCA is very helpful in refecting and expressing the overall situation of the variables under analysis and the total control of  8 International Journal of Analytical Chemistry variables by the researcher, which helps to identify and eliminate problematic samples and abnormal variables, thus improving the accuracy and precision of the mathematical analysis model [18][19][20].Te use of PCA can simplify complex multivariate data systems, and more studies have reported its use in the study of TCM.Many studies have been reported on its use in the study of Chinese medicine and natural drugs [21][22][23].In this study, the results of PCA showed that the order of infuence of 10 components in the classifcation of raw and salt-processed RD was LA > DB > Discriminant analysis is a supervised classifcation technique belonging to chemometrics, which classifes certain objects studied based on certain observed indicators.In the quality control experiment of TCM, discriminant analysis can establish a discriminant based on the observation data of a batch of known samples of various types and then classify the unknown types of samples.FDA is a common method in discriminant analysis, which is generally used to discriminate two kinds of quantitative data [24].It uses the idea of onedimensional ANOVA to reduce the sample points in the ndimensional space to one-dimensional data by means of linear functions and then classifes the sample points to be judged into diferent categories according to the distance between samples.FDA can make the diferences between sample points in the same category as small as possible and make the diferences between sample points in diferent categories as large as possible, thus efectively improving the discriminant efciency [25,26].Tis analytical approach of the FDA was applied in this paper, and the results showed that the model and algorithm given in the paper were efective and useful for the classifcation of raw and salt-processed RD.

Conclusion
In this study, the content of 10 chemical components before and after processing with salt of RD was determined by HPLC-DAD and was sorted through PCA to rank the importance of each chemical component.At the same time, the samples were classifed and verifed by the FDA.Te results showed that the components in RD after processing with salt had internal transformation, the contents of LA, CaA, DA, and AVI increased, and the contents of CA, LN, IB, IA, IC, and DB decreased.In the classifcation of raw and saltprocessed RD, the order of importance of each chemical component was LA > DB > IA > IC > IB > LN > CA > DA > AVI > CaA.Tese components could be used as diferential components to identify raw and salt-processed RD.Tis study provides a comprehensive and quantitative chemical pattern recognition and quality evaluation method for the identifcation of TCM before and after processing.Tis method could also provide a scientifc basis for further research on the spectrum-efect relationship and mechanism of action.

Figure 4 :
Figure 4: Changes in content of RD after processing with salt.

Figure 6 :
Figure 6: FDA graph of raw and salt-processed RD.

Table 1 :
Te sample information of raw and salt-processed RD.

Table 4 :
Te similarity results of 13 batches of RRD.

Table 5 :
Te similarity results of 13 batches of SRD.

Table 8 :
Principal component eigenvalue and contribution rate.

Table 9 :
Principal component loading matrix.value of the load of the compound in the "most informative" frst principal component, it could be seen that the importance of the above 10 compounds in the quality control of raw and salt-processed RD was LA> DB > IA > IC > IB > LN > CA > DA > AVI > CaA.4.4.2.Fisher Discriminant Analysis.FDA is one of the methods of discriminant analysis.It uses the idea of variance analysis to project points in a high-dimensional space to a low-dimensional space to construct one or more linear discriminant functions in diferent-dimensional spaces.Te contents of 10 compounds in RRD and SRD were selected as variables to generate FDA.Te results of the coefcients are shown in Table10, X 1 , X 2 , . .., X 10 in the function expressions were used to represent the normalized data of the content of each compound respectively, and the function expressions were shown as follows: RRD � 323.095X 1 + 490.050X 2 + 2368.596X 3

Table 10 :
Fisher linear discriminant function coefcients of raw and salt-processed RD.