Qi Junsheng, Feng Shenping^#, Yang Xiangliang^##, Lu Xiaohua^##, Zhou Jingyan^##, Xu Huibi^##, Guan Jinghua^###
(Department of Chemistry, Chongqing Three Gorges College, Wan Zhou City,404000; ^#Southest China Normal University, Chongqing ,400715; ^##Department of Chemistry, Huazhong University of Science and Technology, Wuhan,430074; ^###The First Hospital in Wuhan, Wuhan,430022)

Recerived Nov. 5, 2000.

Abstract The relationship between drug's characteristics and the contents of trace elements in Chinese medicinal herbs was studied. 42 kinds of trace elements in 105 Chinese medicinal herbs were determined. Factor analysis were applied to study the distribution characteristics of 42 kinds of trace elements in 105 Chinese medicinal herbs. The results of factor analysis showed that a ten-factor model interprets the correlation of these trace elements. The result of Q-type cluster analysis showed that the samples can be clustered reasonably into different groups. According to the traditional four characteristics of these medicinal herbs, the accuracy of classification is 78.1%. It was certified that the amount of trace elements is one of the main factors determining the four characteristics in Chinese medicinal herbs. There is a significant correlation between the curative effects and the amount of the trace elements in Chinese medicinal herbs.
Keywords Trace elements, Four Characteristics of herbs, Factor analysis, Cluster analysis

1. INTRODUCTION
Four characteristics of Chinese medicinal herbs has been used to instruct use of drugs in clinic of traditional Chinese medicines, which is one of the core theories of Chinese medicine and pharmacology, but, its mechanism has not been interpreted up to date. It can be said certainly there is a relationship between the characters of traditional Chinese medicine and the chemical compositions. The chemical compositions of traditional Chinese medicines are quite complex. Generally, the organic compositions determine the factors affecting the function and potency of traditional Chinese medicines.
In recent years, Guan Jinghuan et al^[1-5] discovered that trace elements (including major elements) bridge the message between animals and plants, which is one of the main factors determining the four characteristics of Chinese medicinal herbs. Therefore, the research of trace elements in Chinese medicinal herbs is one of the methods to interpret the mechanism of characters and functions of traditional Chinese medicines.
    Various chemometric methods have been used in analyzing large data sets obtained in chemical and medical measurement. Factor analysis and cluster analysis^[6-9] are useful for this purpose in this work, 42 kinds of trace elements in 105 different Chinese medicinal herbs were determined. Factor analysis and cluster analysis were used to study the association of these trace elements. As a result, it was certified that the amount of trace elements was one of the main factors determining the four characteristics in Chinese medicinal herbs.

2. EXPERIMENTAL
2.1 Sampling and determination of the trace elements
05 Chinese medicinal herbs and the place of origin are omitted. Analytical methods of 42 kinds of trace elements are listed in Table 1

Table 1 Determined methods of trace elements

2.2 Factor analysis
Original data sets were first transformed logarithmically. The factor analysis computer program in the advanced statistical package SPSS/PC^+[10,11] was used to analyze the data sets to study the association of the trace elements.
2.3 Cluster analysis
Using the logarithmically transformed data sets, the computer program for cluster analysis in the above mentioned package was used to classify 105 samples of Chinese medicinal herbs. The Euclidean distances was used as the measurement of the similarity of sample and the method of average linkage between groups was used to combine the samples.

3. RESULTS AND DISCUSSION
3.1 The concentrations of trace elements  
The concentrations of 42 kinds of trace elements in Gastrodia Tuber are listed in Table2 and the others are omitted in this article.

Table 2 Concentrations of trace elements in Gastrodia Tuber (mg/g)

3.2 Factor analysis
3.2.1 Correlation between the variables   
Table 3 is the part of correlation matrix of 42 variables. Nearly 70% of the correlation coefficient in correlation matrix are over 0.3. All variables have large correlation with at least one other variables. The matrix is therefore appropriate for factor analysis.

Table 3 Part of the correlation matrix of 42 variables

3.2.2 Factor analysis  
Table 4 is the initial statistics of factor analysis. The variance explained by each factor is listed in the column labeled Eigenvalue. The next column is the percentage of the variance attributable to each factor. The last column is the cumulative percentage.

Table 4 Initial Statistics

The goal of the factor extraction step is to determine the number of principal factors. In this paper, principal component analysis was used to obtain estimates of the initial factors. The initial statistics from principal component analysis of the correlation matrix are listed in Table 4. Table 4 shows that almost 81% of the total variances is attributed to the first ten factors, and the remaining thirty-two factors together account for only 19% of the variances. Thus, a model with ten factors may be adequate to represent the data. From the estimation of the factor number of ten, the factor matrix associating the variables and factors was obtained, the coefficient in which are also the correlation between the factors and variables. Factors with large coefficients (in absolute value) for a variable are closely related to the variable. Although the factor matrix in Table 5 (only several data sets are listed) indicates the relationship between the factors and the individual variable, it is difficult to identify meaningful factor in the matrix. The rotation phase of factor analysis is an attempt to transform the initial factor matrix into a simple one that is easier to interpret. A variety of algorithms is used for transformation different rotation methods may result in the identification of somewhat different factors. The varimax method was used in the present work. Some elements of the rotated factor matrix are shown in Table 6.

Table 5 Factor Matrix PC Extracted 10 factors

Table 6 Part Factors of the Rotated Factor Matrix

3.2.3 Interpretation of factors    
One of the goals of factor analysis is to identify factors that are substantively meaningful after rotation, interpretation of the factors(in Table 6) appears possible. For example, the first factor shows a high positive correlation with variables Be, Al, Si, V, Fe, Co, Y, La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, and the second factor a high correlation with variables Mn, Zn, Ni and Cd. These two factors show that these trace elements have a strong tendency forming coordination compounds with organic components in Chinese medicinal herbs. It could be the coordination compounds that cure the sickness to save the patients. Although Table 2 shows that the contents of rare earth elements in Chinese medicinal herbs are below l mg/g, there are significant correlation between them and Chinese medicinal herbs. Some research work on rare earth elements in Chinese medicinal herbs have been reported. Mr Fang Yu^[12] studied the relationship between rare earth elements' ions(La³⁺, Nd³⁺,Sm³⁺,Yb³⁺)and organic components (baicalin). The relationship between rare earth elements ions (La³⁺,Nd³⁺,Sm³⁺, Yb³⁺)and organic components(baicalein) was also studied potentiometrically by Sun Zuomin^[13].It showed that baicalein has very strong capability to form coordination compounds with rare earth elements. Therefore, study of the relationship between rare earth elements and Chinese medicine and pharmacology may reveal the treatment mechanism of Chinese medicinal herbs. The first and the second factors are more important than others in the ten factors. In order to further study these variables, the factor analysis was used to study the association of 15 kinds of rare earth elements in 105 Chinese medicinal herbs. The results of the factor analysis showed that a model with two factors may be adequate to represent the data. The first factor is highly correlated with variables La,Ce,Pr,Nd,Sm,Eu and factor2 with variables Y,Tb,Dy,Ho,Er,Tm,Yb,Lu. Variables associated with the first factor are light rare earth elements, and variables associated with the second factor are heavy rare earth elements, which showed that the data sets has excellent classification ability according to the traditional method. Both light and heavy rare earth elements decide the different effects of Chinese medicinal herbs. In order to illustrate light and heavy rare earth elements in Chinese medicinal herbs again, factor analysis was used to study the association of the elements of the first and the second factors in previous ten factors. Rotated factor matrix of 19 variables is shown in Tables 7. The first factor shows a high positive correlation with variables V,Fe ,Co,Ni,La,Ce, Pr,Nd,Sm,Eu. It shows that light rare earth elements forming coordination compounds are similar to Fe,V,Co,Ni.

Table 7 Rotated Factor Matrix of 19 Variables

Authors inferred that light rare earth elements may be essential trace elements in human body. Therefore, the total amount of trace elements determined reveals inner messages of Chinese medicinal herbs. It can also be used to study the efficient components and the mechanism of curing illness of Chinese medicinal herbs.

3.3 Cluster analysis  
In cluster analysis, the pretreatment of data, the measurement of similarities and the criteria for combining variables into clusters should be considered. In this work the logarithmically transformed data sets were used for cluster analysis as well, and the method of average linkage between groups was used. Q-type cluster analysis was used for classification of 105 Chinese medicinal herbs samples. Table 8 (only several data sets are listed)is the agglomeration schedule and Figure 1 is the dendrogram of the cluster analysis of the 105 samples.

Table 7 Agglomeration Schedule using Average Linkage (Between Groups)

Figure 1 Dendrogram using Average Linkage(Between Groups)

The results of Q-type cluster analysis show that through using the forty-two trace elements as characteristics variables 105 Chinese medicinal herbs can be classified correctly into four groups. Drugs of neutral nature usually tend to be either slightly hot or slightly cold, which are correct in any groups. The accuracy of classification is 78.1%(See Table 9)

Table 9 A comparison between classification results by cluster analysis of 105 Chinese medicinal herbs and the ones according to the four characteristics

Summarily, both factor analysis and cluster analysis certified that there are correlation between trace elements and characteristics of Chinese medicinal herbs. It provides some guidance for further studying efficient components of Chinese medicinal herbs. There may be two kinds of functions for trace elements to act on organism: one is to immediately act on organism and give rein to property of a medicine; the other is to form coordination compounds with organic molecules to increase the drugs' property. Thus, there are apparently curative effect to whole prescription of Chinese medicines, but no apparently curative effect to certain drug of the whole prescription of Chinese medicines, or even if there is apparently curative effect to certain Chinese medicine, activity may lose^[14] if its organic components are extracted and purified.

REFERENCES    
[1] Guan Jinghuan. Acta Medica Sinica,1990, 5 (5): 40.
[2] Tang Xuejun, Guan Jinghuan. Trace Elements and Healthy study, 1994, 11 (4): 24.
[3] Qi Junsheng, Xu Huibi, Zhou Jingyan. World Elemental Medicine, 1997,4(3):35.
[4] Qi Junsheng, Xu Huibi, Zhou Jingyan. Journal of Central China Normal University, 1997, (special issue): 197.
[5] Qi Junsheng, Xu Huibi, Zhou Jingyan. Chinese Journal Analytical Chemistry, 1998, 26 (11): 1309.
[6] Lu Xiaohua.Chemometrics .Wuhan:Huazhong University of Science and Technology (HUST)Press, 1997.
[7] Lu Xiaohua. Fuel, 1995, 74 (9): 1381.
[8] Malinowski E R, Howery D G. Factor Analysis in Chemistry. New York: Wiley-Interseiene, 1980.
[9] Lin S G, Yuan P J, Shen D X. Multivariate Statistics. Wuhan: HUST Press, 1987.
[10] Norusis M J. "Advance statistics SPSS/PC+", SPSS Inc, 1986.
[11] Hull Nie N H, Jenkins C H. J.G.Statistical Package for the Social Sciences. 2nd Edn, New York: McGraw-Hill, 1975.
[12] Fang Yu. Chinese Biochemical Journal, 1991, 7 (6): 753.
[13] Cao Zhiquan. Trace Elements and Chinese Medicine and Pharmacology. Beijing: Chinese Medicine and Pharmacology Press,1993.
[14] Wang Aifang. Bulletin of Pharmaceutical Sinica, 1981, 16 (3): 61.

[ Back ] [ Up ] [ Next ] Mirror Site in USA  Europe  China  CSTNet ChinaNet