Zhou 2015

Journal of Pharmaceutical and Biomedical Analysis 111 (2015) 16

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

j o ur na l ho mepage: www.elsev ier .com/ locate / jpba

Chemometric analyses for the characterization oseeds of Descurainia sophia (L.) based on HPLC n

Xidan Zhoua, Liying Tanga, Hongwei Wua, Guohong Zhoua, TinShunxin Lib, Zhuju Wanga,

a Institute of Chinese Materia Medica, China Academy of Chinese Medical Science, Beijing 100700, Chinab Department o

a r t i c l

Article history:Received 6 DeReceived in reAccepted 8 MaAvailable onlin

Keywords:Descurainia soHigh-performwith diode arrMultivariate sProcessing mechanismPlantago depressa Willd

hort fmany

DSS s whreforwas ee diffeecha

1. Introduction

Descuraito family Brfor the treamiddle Asiaexploited tomote urinamost Chinechallenges oview of the ants, it seemor microscoaccurately a

Currentlimportant r

Abbreviatio(L.); HCA, hierDA, partial leaWilld.; RDSS, tmedicine.

Corresponof Chinese MedChina. Tel.: +8

E-mail add

which can systematically characterize the constituents of samples

http://dx.doi.o0731-7085/ nia sophia (L.) Webb ex Prantl, an annual dicot belongingassicaceae (Cruciferae), has been used in folk medicinestment of throat diseases, measles and smallpox in the

[1,2]. In China, its seeds (DSS) have been extensively relieve cough, prevent asthma, reduce edema and pro-tion for thousands of years [3,4]. However, parallel tose medicinal herbs, DSS have been encountering thef varietal complexity and counterfeit incorporation. In

morphological specicity and high similarity of adulter-s inaccessible that either morphological characteristicspic characteristics could be able to distinguish themnd handily.y, the chromatographic ngerprint technique plays anole in the quality control of Chinese medicinal herbs,

ns: DAD, diode array detector; DSS, the seeds of Descurainia sophiaarchical cluster analysis; PCA, principle components analysis; PLS-st squares discriminant analysis; PDS, the seeds of Plantago depressahe roasted seeds of Descurainia sophia (L.); TCM, traditional Chinese

ding author at: Institute of Chinese Materia Medica, China Academyical Science, No. 16 Nanxiaojie, Dongzhimennei Ave., Beijing 100700,

6 10 64033301; fax: +86 10 64033301.ress: [email protected] (Z. Wang).

and focus on the identication and assessment of the components[5]. Due to the complexity and world-wide applicability of Chineseherbs, chromatographic ngerprinting has been accepted interna-tionally as a strategy to assess the quality of such substances [6].In this paper, HPLC-DAD was applied to construct the chromato-graphic ngerprints.

Processing is a characteristic pharmaceutical skill which sig-nicantly discriminates western medicine from Chinese medicine.Proper processing may remove or reduce the toxicity, drastic prop-erties and side effects, promote therapeutic effects and modify thenature and action. Hence, in order to guarantee the safety and effec-tiveness, it is signicant and necessary to control and regulate thequality of processed products. As for DSS, due to the drastic effectsand potential side-effects in the case of raw DSS, processed DSShave been dominating in clinical application [7]. As described inthe textbook, the raw DSS are usually indicated for ascites andhyposarca while the roasted DSS are mainly used for cough anddyspnea because roasting can moderate the extremely cold natureand protect lung from drastic damage [8]. Thus, it is necessary todiscriminate raw and roasted DSS in clinical practice. Recently,chemometrics have been increasingly applied to the botanicaland geographical characterization and authentication of food andmedicine [918]. In this study, HPLC ngerprints of the raw androasted DSS were compared, and the ngerprint data sets weresubmitted for classication to several chemometrics methods, such

rg/10.1016/j.jpba.2015.03.0102015 Elsevier B.V. All rights reserved.f Chinese Medicine, Nanyang Medical College, Nanyang, Henan 473000, China

e i n f o

cember 2014vised form 5 March 2015rch 2015e 17 March 2015

phia (L.)ance liquid chromatographyay detector ngerprintstatistical analysis

a b s t r a c t

The seeds of Descurainia sophia (L.) (sChina, have attracted the attention of present study, the raw and processedon HPLC ngerprints. Moreover, peakraw and roasted DSS were found. Thewith multivariate statistical analysis which were mainly responsible for thlight on illustrating the processing midentication of authenticity.f raw and processedgerprints

g Wanga, Zhenzhen Koua,

or DSS below), with a long history of medicinal utilization in Chinese medicine practitioners for the potent efcacy. In thewere differentiated by several chemometrics methods basedich were mainly responsible for the differentiation betweene, the method of the chromatographic ngerprints combinedffective and reasonable in orientating chemical constituentsrentiation between raw and roasted materials, thus sheddingnism. Whats more, this method can also be applied in the

2015 Elsevier B.V. All rights reserved.

2 X. Zhou et al. / Journal of Pharmaceutical and Biomedical Analysis 111 (2015) 16

as principle components analysis (PCA), hierarchical cluster analy-sis (HCA) and partial least squares discriminant analysis (PLS-DA).According to the statistical results combined with the chromato-graphic ngerprints, peaks responsible for discrimination betweenraw and prcourse of pr

As mentmedicinal mother mediIn this studWilld. (PDSHCA was alto explore adulteratindifferent pr

2. Materia

2.1. Materi

21 batchmedicinal hlocated in BProvince (SChen, DepaThen, RDSS(chosen ranprocessing tion, 10 batChinese herbeen foundcation andon their diftics, it turnewhile the rsamples weminimize aroasted vouMateria Me

PhosphoSinopharm tonitrile anThermoFishwas lteredEquipment

2.2. Sample

The testefor 1 min, aple powder80% methanextract, themembrane,analysis. Thfor HPLC asamples sta

2.3. Chrom

All HPLseries (Shana LPG-3400an injectorment, a DADworkstation

(4.6 mm 250 mm, 5 m) (Shanghai, China) was used. The mobilephase consisted of 1% phosphoric acid (A) and acetonitrile (B).The gradient program was developed as follows: 1014% B for012 min, keep 14% B for 1218 min, 1419% B for 1830 min, stay

for 3 min,

min.raturd th

ethod

testsed aslicatiy. Siallel d byratur

ata a

C ch autosing.tion s) (Co

wasmonhenmmoativeted sr statnd Pm.

ults

ptimi

ordetion tion)nol, 81 h, 1e rean uso folventitionted crepa

h.

ptimi

bile nolan, wc behtter ns (fred, l to iiffereocessed DSS were found and how they change in theocessing was also analyzed.ioned above, there are plenty of adulterants of DSS inarkets. Meanwhile, DSS are mixed with the seeds of

cinal herbs as a result of the huge price difference [19].y, HPLC ngerprints from the seeds of Plantago depressa) were also obtained to compare with that of DSS, andso conducted to differentiate them. Moreover, in orderthe inuence on the quality of PDS caused by varyingg degree, chromatographic ngerprints of mixtures ofoportions of DSS and PDS were performed.

ls and methods

als and reagents

es of raw DSS were purchased from different Chineseerbs suppliers in the Chinese herbal medicine marketozhou, Anhui Province (Samples 19) and Anguo, Hebeiamples 1021), and authenticated by professor Suiqingrtment of Pharmacognosy of Henan University of TCM.

(Samples 2235) were produced by roasting raw DSSdomly from 21 batches of raw samples) according to themethod described in Chinese Pharmacopeia. In addi-ches of PDS (Samples 3645) were obtained from thebal medicine market in Anguo where these seeds have

occasionally mixed with DSS. After the careful identi- authentication of these 10 batches of samples basedference in morphological and microscopic characteris-d out that only samples 38, 44, 45 were unadulteratedemaining were more or less mingled with DSS. All there stored in a dry ambience at constant temperature tony changes through degradation, and all the raw andcher specimens were deposited in institute of Chinesedica, China Academy of Chinese Medical Science.ric acid (guaranteed reagent) was purchased fromChemical Reagent Co., Ltd. (Shanghai, China). Ace-d methanol for HPLC analysis were supplied byer Scientic Inc. (Shanghai, China). The water used

with the solvent lter (Tianjin Jinteng Experiment Co., Ltd., Tianjing, China).

preparation

d samples were crushed into powder with a pulverizernd passed through a 65 m-mesh sieve. Each sam-

(1.0 g) was weighed accurately and reuxed in 20 mLol solution for 30 min. The residue was sifted from then this solution was ltered with a 0.45 m microporous

and a 5 L aliquot of the ltrate was injected for HPLCe extract solutions of all the samples were summitednalysis within 24 h after preparation to maintain thebility.

atographic conditions

C analyses were performed with a Dionex U-3000ghai, China) equipped with a SR-3000 Solvent Rack,SDN Quaternary Pump, a WPS-3000SL Auto sampler,

with a 100 L loop, a TCC-3000RS Column compart--3000RS detector and Chromeleon 7 chromatography. A Thermo Scientic Acclaim TM 120-C18 column

19% B 40505155tempe5 L an

2.4. M

All preparby repper dain parassessetempe

2.5. D

HPLgratedprocesevalua2004 Asampleones aplate. Tand coThe relcalculafurthe16.0, aprogra

3. Res

3.1. O

In extracextracmetha(0.5 h, test. Thtive thwas altion soIn addextracwere pfor 0.5

3.2. O

Momethasolutiographiwas besolutiocompahelpfuLast, d035 min, 1925% B for 3540 min, maintain 25% B for 2540% B for 5051 min, and nally hold 40% B for

The ow rate was kept at 1.0 mL/min and the columne was maintained at 30 C. The injection volume wase detective wavelength was selected at 330 nm.

ology validation

below were carried out on the raw DSS extract solutions described in Section 2.2. The precision was determinedng HPLC injections of the same sample solution 6 timesx independent samples were extracted and analyzedfor the evaluation of repeatability. The stability was

measuring a single sample solution stored at roome for 0 h, 2 h, 4 h, 8 h, 16 h, and 24 h.

nalysis

romatographic data of the 45 tested samples were inte-matically and exported as *.AIA format les for further

First, the *.AIA les were imported into the similarityystem for chromatographic ngerprint of TCM (Versionmmittee for the Pharmacopeia of PR China.). A reference

selected stochastically from the most representationalg the middle of analysis sequence to generate a tem-, all of the samples were overlaid based on the template,n peaks were rstly aligned to the ones in the template.

retention time (RRT) and relative peak area (RPA) wereimultaneously and can be exported as an excel le foristical analysis. HCA and PCA were obtained by SPSSLS-DA was conducted by the SIMCA-p 13.0 software

and discussion

zation of extraction conditions

r to obtain satisfactory extraction efciency, themethods (reuxing, ultrasonic and cold-macerating, extraction solvents (20% methanol, 40% methanol, 60%0% methanol and 100% methanol) and extraction time.5 h, 2 h, and 2.5 h) were optimized by using univariatesult indicated that reuxing extraction was more effec-ltrasonic extraction and cold-macerating extraction. Itund that 80% methanol was the most efcient extrac-

among the tested different concentrations of methanol., it was demonstrated that most components could beompletely within 0.5 h. Finally, the sample solutionsred by reuxing extraction with 20 mL 80% methanol

zation of chromatographic conditions

phases, such as methanolwater, acetonitrilewater,cid aqueous solution, and acetonitrileacid aqueousere examined and compared to obtain good chromato-avior. As a result, acetonitrileacid aqueous solutionthan others. Moreover, different kinds of acid aqueousormic acid, acetic acid and phosphoric acid) were alsoit was demonstrated that phosphoric acid was moremprove the peak shape and resolution of many peaks.nt concentrations of phosphoric acid (0.2%, 0.5%, 0.8%,

X. Zhou et al. / Journal of Pharmaceutical and Biomedical Analysis 111 (2015) 16 3

Fig. 1. The chromatographic ngerprints of DSS and PDS samples: (A) DSS; (B) RDSS; (C) PDS sample 43 and (D) PDS sample 45.

and 1%) wecondition w

3.3. Method

RRT andthe estimatresults wertions (RSD)2.69% respetively; and all results inHPLC-DAD

PLC

atchorres

nm, rintsatogrl peappea, 24en itre compared, and the data indicated that the optimalas acetonitrile1% phosphoric acid aqueous solution.

ology validation

RPA of ten characteristic peaks were calculated forion of precision, repeatability and stability, and thee as follows: precisionthe relative standard devia-

of RRT and RPA were found not to exceed 0.41% andctively; repeatabilitybelow 0.19% and 2.77% respec-stabilityless than 0.27% and 4.04% respectively. Thus,

3.4. H

45 btheir cat 330ngerpchromsevera41 disaas 4, 22

Wh

dicated that the quality of the studied samples and themeasurements were stable and under control.

was appare44, 45 were

Fig. 2. (A) HCA of DSS and RDSS samples and (B) HCA ofngerprints of DSS, RDSS and PDS

es of samples included 21 DSS, 14 RDSS and 10 PDS, andponding chromatographic ngerprints were recordedwhich were showed in Fig. 1. For the chromatographic

of DSS and RDSS, it can be intuitively seen that manyaphic peaks changed to varying degrees. For example,ks (25, 21) areas declined obviously, and even peaks 28,red after processing. On the contrary, many peaks, such, 26, 32 and 33, were present in the RDSS.

comes to the chromatographic ngerprints of 10 PDS, it

nt that the chromatographic ngerprints of samples 38,

different from that of the remaining samples because

DSS and PDS samples.

4 X. Zhou et al. / Journal of Pharmaceutical and Biomedical Analysis 111 (2015) 16

of the great variance of the existing conditions of ve main peaks(13, 16, 18, 20, 25) located in 717 min, which were precisely thecharacteristic and common peaks of DSS. In accordance with theidentication result above, chromatographic ngerprints also pro-vided more accurate and intuitive evidence.

3.5. HCA for identication of DSS and RDSS, DSS and PDS

HCA is a means of structuring a complex set of observationsinto unique, mutually exclusive groups (clusters) of subjects sim-ilar to each other with respect to certain characteristics [20,21].HCA results of DSS and RDSS were showed in Fig. 2A. In a whole, itwas evident that DSS and RDSS samples were clearly clustered intotwo groups, which means that the processing procedures causedchanges in the composition and/or content of components in DSS.

As for DSS and PDS, the HCA result (Fig. 2B) also conrmed theobvious difference obtained by the chromatographic ngerprintsdirectly, which perspicuously classied 10 PDS samples into twogroups: 38, 44, 45 and the remaining samples.

3.6. PCA for identication of DSS and RDSS

Specically, PCA takes a data matrix of n objects by p variables,which may be correlated, and summarizes it by uncorrelated axes(principal components) that are linear combinations of the originalp variables [22,23]. To achieve a balance between clarity of repre-sentation and oversimplication, 52 peaks (variables), which werepresent in at least 10 samples, were selected to conduct statisticalanalysis. The PC1 versus PC2 biplot (Fig. 3) accounted for 57.38%data variance (PC1 = 45.80%, PC2 = 11.58%), and two clusters of DSSand RDSS were identied. The RDSS objects with relatively high

Fig. 3. PCA biplots PC1-PC2 (57.38% data variance explained) of DSS and RDSS sam-ples (the red circles represent 52 variables; the blue circles stand for 35 samples).(For interpretation of the references to color in this gure legend, the reader isreferred to the web version of this article.)

positive scores on PC1, grouped together fairly tightly, while mostDSS objects bunched together with negative PC1 scores. On PC2, thetwo groups both had positive and negative scores. Thus, DSS werediscriminated from RDSS on PC1. According to the contributions toPC1 of all variables, it can be easily seen that many variables wereresponsible for the composition of PC1, among them peaks 4, 26,24, 22 and 25 featured strongly in identifying DSS and RDSS.

Fig. 4. (A) PLS DSS and RDSS samples; (C) VIP (variable importance for the project) plot ofPLS-DA of DSS-DA score scatter plot of DSS and RDSS samples; (B) PLS-DA loading scatter plot of and RDSS samples.

X. Zhou et al. / Journal of Pharmaceutical and Biomedical Analysis 111 (2015) 16 5

Fig. 5. Chrom DS; S2S5: 40% DSS + 6 : 80%

3.7. PLS-DA

As a supdifference aresponsibleations withcomponentDA was alsoand RDSS whow 35 obsshowed theother patte

Moreovewhich dispthe dummyvicinity of tbetween thto the Y variwere strong

In orderdiscriminat(Fig. 4C) wavariables boplot, many vthese variabAmong theming consisteR2Y = 0.989the data an

3.8. The estmixtures of

Althoughnating DSSlimited bectication. Tdifferent pe90%. As it is clustered todistributed

orateecoms w

fromssaryf difg adtely.

clus

ummPLS-Dinat

sultsplayeed af

it plly, srly, tatographic ngerprints of PDS mixed with different percentages of DSS (S1: 100% P0% PDS; S6: 50% DSS + 50% PDS; S7: 60% DSS + 40% PDS; S8: 70% DSS + 30% PDS; S9

for identication of DSS and RDSS

ervised recognition pattern, PLS-DA can maximize themong the groups and aid in the screening of the markers

for class separation rather than explaining the vari-in a data set [24,25]. In order to nd the potentials for the discrimination between DSS and RDSS, PLS-

performed. First, a reasonably good separation of DSSas obtained in the scatter plot (Fig. 4A), which displayedervations were situated with respect to each other and

possible presence of outliers, groups, similarities andrns in the data.r, the loading scatter plot (Fig. 4B) was also conducted,layed the relation between the X-variables (52) and

Y-variables (2). By default, X-variables situated in thehe Y-variables have the highest discriminatory powere classes. As showed, variables 25, 4, 26, 24, stand closeables and far from the origin, guring that such variablesly responsible for discrimination.

to weigh the effect of importance of every variable onion, the VIP (variable importance for the project) plot

incorpually bsampleeffectsis necetures ovaryinaccura

4. Con

In svised discrimThe re22, 32 emergmakesmentaSimilas performed, which summarized the importance of theth to explain X and to correlate to Y. According to the VIPariables had VIP-values larger than 1, which meant thatles were primarily responsible for the discrimination., 26, 24, 4, 25, 32, 22 had the largest VIP-values, keep-nt with the analytical result above. With R2X = 0.528,, Q2 = 0.976, the PLS-DA model was demonstrated to td predict new data well.

ablishment of chromatographic ngerprints ofdifferent proportions of DSS and PDS

microscopic identication was efcient in discrimi- from PDS, the application of the technique has stillause of its failure to provide valid information on quan-hus, chromatographic ngerprints of PDS mixed withrcentages of DSS were performed, ranging from 10% toseen from Fig. 5, the chromatographic peaks of PDS weregether after 30 min, while the peaks of DSS were mainlyin 520 min range. When more than 30% DSS was

authenticitple more efof morpholidenticatiople, sensitivChinese me

Acknowled

The authcation of al

References

[1] R.E. Blackrainia sop817824

[2] N.P. Bekkseeds, Ch

[3] K. Sun, Xthe seeds: 10% DSS + 90% PDS; S3: 20% DSS + 80% PDS; S4: 30% DSS + 70% PDS;DSS + 20% PDS; S10: 10% DSS + 90% PDS; S11: 100% DSS).

d into PDS, the front peaks belonging to DSS have grad-e the main and strong peaks. Consequently, these

ere likely to produce completely different medicinal those pure PDS when applied in clinic. As a result, it

to establish the chromatographic ngerprints of mix-ferent proportions of DSS and PDS. That will provideulterating degree for reference more intuitively and

ion

ary, whether the unsupervised HCA and PCA or super-A, both proved to be satisfactory for matching anding the chromatographic ngerprints of DSS and RDSS.

of PCA and PLS-DA indicated that peaks 26, 24, 4, 25,d dominating roles. Obviously, many new componentster roasting. The accurate orientation of these peaksossible to explain the dissimilarity of efcacy funda-hining a spotlight for revealing processing mechanism.his method can also be applied in the identication of

y, which can detect the divergence among every sam-fectively and precisely, making up for the subjectivityogical identication and inconvenience of microscopicn. The developed method was demonstrated to be sim-e and reproducible, and will be widely adapted to otherdicinal herbs.

gments

ors are grateful to professor Suiqing Chen for authenti-l the samples.

shaw, Control of stinkweed (Thlaspi arvense) and ixweed (Descu-hia) in winter wheat (Triticum aestivum), Can. J. Plant Sci. 70 (1990).er, N.T. Ulchenko, A.I. Glushenkova, Lipids from Descurainia sophiaem. Nat. Compd. 4 (2005) 346347.. Li, J.M. Liu, J.H. Wang, W. Li, Y. Sha, A novel sulphur glycoside from

of Descurainia sophia (L.), J. Asian Nat. Prod. Res. 7 (2005) 853856.

6 X. Zhou et al. / Journal of Pharmaceutical and Biomedical Analysis 111 (2015) 16

[4] D.C. Tang, Processing of Chinese medicinal herbs, in: Science of Chinese Mate-ria Medica, Publishing House of Shanghai University of Traditional ChineseMedicine, Shanghai, 2003.

[5] C. Tistaert, B. Dejaegher, Y. Heyden, Chromatographic separation techniquesand data handling methods for herbal ngerprints: a review, Anal. Chim. Acta690 (2011) 148161.

[6] Y.N. Ni, Y.H. Lai, S. Brandes, S. Kokot, Multi-wavelength HPLC ngerprints fromcomplex substances: an exploratory chemometrics study of the Cassia seedexample, Anal. Chim. Acta 647 (2009) 149158.

[7] W.H. Lv, M.F. Ma, H.Y. Jiang, Evolution and exploration of application ofprocessing of the seeds of Descurainia sophia (L.), Chin. Tradit. Pat. Med. 28(2006) 517519.

[8] Q.F. Gong, Science of Processing Chinese Materia Medica, Chinese Press of Tra-ditional Chinese Medicine, Beijing, 2003.

[9] I.K. Karabagias, A.V. Badeka, S. Kontakos, S. Karabournioti, M.G. Kontominas,Botanical discrimination of Greek unioral honeys with physico-chemical andchemometric analyses, Food Chem. 165 (2014) 181190.

[10] S.A. Drivelos, K. Higgins, J.H. Kalivas, S.A. Haroutounian, C.A. Georgiou, Datafusion for food authentication. Combining rare earth elements and trace metalsto discriminate Fava santorinis from other yellow split peas using chemometrictools, Food Chem. 165 (2014) 316322.

[11] A. Palacios-Morillo, J.M. Jurado, A. Alcazar, F.D. Pablos, Geographical charac-terization of Spanish PDO paprika by multivariate analysis of multielementalcontent, Talanta 128 (2014) 1522.

[12] R. Moreno-Rojas, P.J. Sanchez-Segarra, F. Camara-Martos, M.A. Amaro-Lopez,Multivariate analysis techniques as tools for categorization of Southern Spanishcheeses: nutritional composition and mineral content, Eur. Food Res. Technol.231 (2010) 841851.

[13] K. Chudzynski, J. Falandysz, Multivariate analysis of elements content of LarchBolete (Suillus grevillei) mushroom, Chemosphere 73 (2008) 12301239.

[14] S. Husted, B.F. Mikkelsen, J. Jensen, N.E. Nielsen, Elemental ngerprint anal-ysis of barley (Hordeum vulgare) using inductively coupled plasma mass

spectrometry, isotope-ratio mass spectrometry, and multivariate statistics,Anal. Bioanal. Chem. 378 (2004) 171182.

[15] F. Camin, K. Wietzerbin, A.B. Cortes, G. Haberhauer, M. Lees, G. Versini, Appli-cation of multielement stable isotope ratio analysis to the characterization ofFrench, Italian, and Spanish cheeses, J. Agric. Food Chem. 52 (2004) 65926601.

[16] C.H. Yu, C.Z. Wang, C.J. Zhou, B. Wang, L. Han, C.F. Zhang, X.H. Wu, C.S. Yuan,Adulteration and cultivation region identication of American ginseng usingHPLC coupled with multivariate analysis, J. Pharm. Biomed. Anal. 99 (2014)815.

[17] M.P. La, F. Zhang, S.H. Gao, X.W. Liu, Z.J. Wu, L.N. Sun, X. Tao, W.S. Chen, Con-stituent analysis and quality control of Lamiophlomis rotata by LC-TOF/MS andHPLC-UV, J. Pharm. Biomed. Anal. 102 (2015) 366376.

[18] W.H. Wang, H.J. Dong, R.Y. Yan, H. Li, P.Y. Li, P. Chen, B. Yang, Z.M. Wang, Com-parative study of lanostane-type triterpene acids in different parts of Poria cocos(Schw.) Wolf by UHPLC-Fourier transformation MS and UHPLC-triple quadru-ple MS, J. Pharm. Biomed. Anal. 102 (2015) 203214.

[19] D. Guo, N.N. Chen, Identication of Descurainia sophia and Semen plantaginis byHPCE, China Pharm. 20 (2009) 683684.

[20] K. Ann-Kathrin, G.T. William, W.M. Meinhardt, Hierarchical cluster analysis(HCA) of microorganisms: an assessment of algorithms for resonance Ramanspectra, Appl. Spectrosc. 65 (2011) 165170.

[21] S.K. Yan, W.F. Xin, G.A. Luo, Y.M. Wang, Y.Y. Cheng, An approach to develop two-dimensional ngerprint for the quality control of Qingkailing injection by high-performance liquid chromatography with diode array detection, J. Chromatogr.A 1090 (2005) 9097.

[22] I.E. Frank, J.H. Friedman, A statistical view of some chemometrics regressiontools, Technometrics 35 (1993) 109135.

[23] T. Naes, U. Indahl, A unied description of classical classication methods formulticollinear data, J. Chemom. 12 (1998) 205220.

[24] M. Barker, W. Rayens, PLS for discrimination, J. Chemom. 17 (2003) 166173.[25] S. Bougeard, M. Hana, E.M. Qannari, Continuum redundancyPLS regression:

a simple continuum approach, Comput. Stat. Data Anal. 52 (2008) 36863696.

Chemometric analyses for the characterization of raw and processed seeds of Descurainia sophia (L.) based on HPLC fingerpr...1 Introduction2 Materials and methods2.1 Materials and reagents2.2 Sample preparation2.3 Chromatographic conditions2.4 Methodology validation2.5 Data analysis

3 Results and discussion3.1 Optimization of extraction conditions3.2 Optimization of chromatographic conditions3.3 Methodology validation3.4 HPLC fingerprints of DSS, RDSS and PDS3.5 HCA for identification of DSS and RDSS, DSS and PDS3.6 PCA for identification of DSS and RDSS3.7 PLS-DA for identification of DSS and RDSS3.8 The establishment of chromatographic fingerprints of mixtures of different proportions of DSS and PDS

4 ConclusionAcknowledgmentsReferences