Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
Fourier transform infrared spectroscopy combined with chemometrics for discrimination of Curcuma longa, Curcuma xanthorrhiza and Zingiber cassumunar Eti Rohaeti a, Mohamad Rafi a,b,⇑, Utami Dyah Syafitri c, Rudi Heryanto a,b a
Department of Chemistry, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Jalan Agatis Kampus IPB Dramaga, Bogor 16680, Indonesia Biopharmaca Research Center – Research and Community Empowerment Institute, Bogor Agricultural University, Jalan Taman Kencana No. 3 Kampus IPB Taman Kencana, Bogor 16128, Indonesia c Department of Statistics, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Jalan Meranti Kampus IPB Dramaga, Bogor 16680, Indonesia b
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Discrimination of C. longa, C.
xanthorrhiza and Z. cassumunar by FTIR. Principal component analysis and canonical variate analysis are used for discrimination. Canonical variate analysis gave clearer discrimination between the three species.
a r t i c l e
i n f o
Article history: Received 22 March 2014 Received in revised form 14 July 2014 Accepted 31 August 2014
Keywords: C. longa C. xanthorrhiza Z. cassumunar FTIR Chemometrics Discrimination
a b s t r a c t Turmeric (Curcuma longa), java turmeric (Curcuma xanthorrhiza) and cassumunar ginger (Zingiber cassumunar) are widely used in traditional Indonesian medicines (jamu). They have similar color for their rhizome and possess some similar uses, so it is possible to substitute one for the other. The identification and discrimination of these closely-related plants is a crucial task to ensure the quality of the raw materials. Therefore, an analytical method which is rapid, simple and accurate for discriminating these species using Fourier transform infrared spectroscopy (FTIR) combined with some chemometrics methods was developed. FTIR spectra were acquired in the mid-IR region (4000–400 cm 1). Standard normal variate, first and second order derivative spectra were compared for the spectral data. Principal component analysis (PCA) and canonical variate analysis (CVA) were used for the classification of the three species. Samples could be discriminated by visual analysis of the FTIR spectra by using their marker bands. Discrimination of the three species was also possible through the combination of the pre-processed FTIR spectra with PCA and CVA, in which CVA gave clearer discrimination. Subsequently, the developed method could be used for the identification and discrimination of the three closely-related plant species. Ó 2014 Elsevier B.V. All rights reserved.
⇑ Corresponding author at: Department of Chemistry, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Jalan Agatis Kampus IPB Dramaga, Bogor 16680, Indonesia. Tel./fax: +62 251 8624567. E-mail address: [email protected] (M. Rafi). http://dx.doi.org/10.1016/j.saa.2014.08.139 1386-1425/Ó 2014 Elsevier B.V. All rights reserved.
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Introduction Turmeric (Curcuma longa), java turmeric (Curcuma xanthorrhiza) and cassumunar ginger (Zingiber cassumunar) cultivated mostly in the Java Island are widely used in traditional Indonesian medicines (jamu). C. longa and C. xanthorrhiza are usually aromatic and carminative, and are used to treat stomachic, hepatitis, jaundice, diabetes, atherosclerosis and bacterial infections [1], while Z. cassumunar is usually used to treat various diseases such as asthma, carminative, colic, diarrhea, stomachic, muscle and joint pain [2,3]. The color of their rhizome is yellow to orange for C. longa and C. xanthorrhiza, and pale yellow to yellow for Z. cassumunar. These three plants are belongs to the Zingiberaceae family and therefore, they may have some similar chemical components. Phenolic compounds (diarylheptanoid, diarylpentanoid, phenylpropene derivative) and terpenoids (monoterpenoid, sesquiterpenoid, diterpenoid, and triterpenoid) have been identified in C. longa and some of them also present in C. xanthorrhiza. Diarylheptanoids/curcuminoids (curcumin, demethoxycurcumin, bisdemethoxycurcumin, etc.) and sesquiterpenoids (curcumene, turmerone, zingiberene, etc.) were found as a major group of compounds in C. longa and C. xanthorrhiza [4,5]. While in Z. cassumunar, phenylbutanoids ((E)-1(3,4-dimethylphenyl) butadiene), diarylheptanoids (curcumin), monoterpenoids (sabinene) and sesquiterpenoids (zingiberene) were identified [2]. Curcuminoids are responsible for the color of the three plants used in this study. Due to the similarity of the color of their rhizomes which possess some similar uses, substitution for each other could be possible, especially if in the powdered form and if the price of C. xanthorrhiza is much higher than C. longa or Z. cassumunar. If the substitution occured, it could be a serious problem because of inconsistency in their biological properties. In this case, the identification and discrimination of these closely-related plants are crucial in order to ensure the quality, safety and efficacy of the raw materials before they are converted to final products. Therefore, a rapid, simple and accurate analytical method is essentially required for the discrimination of these species. Several analytical techniques such as spectroscopy (UV–Vis, FTIR, NMR, and MS) and chromatography (TLC, HPLC, and GC) have been used in the development of method for identification and discrimination of medicinal plants. Among these techniques, Fourier transform infrared (FTIR) spectroscopy may be an attractive option because it can meet the criteria of efficient analysis, i.e. easy to use, fast, and inexpensive [6]. Although medicinal plants will contain a large variety of chemical components, their FTIR spectra sometimes are found to differ even in the same species [7]. FTIR has been widely used and is a well-established tool for quality control in various industries, including herbal industry [8]. FTIR spectra contain complex data information describing the overall chemical signal in a sample. Changes in the position and intensity of bands in the FTIR spectra would be associated with the changes in the chemical composition of a sample. Therefore, FTIR spectra could be used to discriminate closely-related species, although the composition of a chemical compound is certainly not known [8]. Obvious advantages of FTIR application to discriminate different medicinal plants are not only effective and specific, but also rapid and non-separative. Discrimination by visual inspection in the FTIR spectra is not easy because the FTIR spectra pattern is very complex, and to address this concern, the aid from chemometrics methods is needed [9]. The advantage of using chemometrics for the interpretation of FTIR is the ability to link the spectral pattern with hidden information contained in a sample [10]. FTIR spectral analysis alone or in combination with chemometrics methods has been extensively used for species identification and discrimination of some closely-related species [11–22]. Because of the advantages
1245
of FTIR spectroscopy, we used this techniques in combination with chemometrics methods for the first time to develop a qualitative method for discrimination of C. Longa, C. xanthorrhiza and Z. Cassumunar. This combined method was successfully applied for the identification and discrimination of the three species.
Material and methods Chemicals Potassium bromide (KBr) for spectroscopy was purchased from Sigma–Aldrich (St Louis, USA). Analytical-grade ethanol for solvent extraction was obtained from Merck (Darmstadt, Germany). Sample preparation Ninety-nine samples consisting of 35 samples of C. longa, 35 samples of C. xanthorrhiza and 29 samples of Z. cassumunar were collected during 2008–2010 from 11 regencies (2–5 samples from each regency) in Java Island, Indonesia: Bogor, Sukabumi, Sumedang, Purworejo, Wonogiri, Karanganyar, Semarang, Kulonprogo, Pacitan, Ponorogo, and Kediri (Table 1). Voucher specimens were deposited at Biopharmaca Research Center, Bogor Agricultural University, Indonesia. All samples were sieved, dried, and pulverized prior to use. There are several methods to obtained IR spectra for the purpose of identification and discrimination as described by Zou et al. [10]. In this experiment, the samples were extracted with a solvent and after evaporating the solvent, the extracts were blended with KBr. This method will provide better resolution in the identification and discrimination of closely related plants [10]. Ethanol was used to extract the samples and after the extraction process, the ethanol was evaporated by rotary evaporator. These ethanol extracts were then used for FTIR measurement. FTIR spectroscopy measurement FTIR spectra were obtained using a Bruker Tensor 37 FTIR spectrophotometer equipped with deuterated triglycine-sulphate (DTGS) as detector and controlled by OPUS 4.2 software (Bruker, Germany). About 5 mg of ethanol extract of each sample was mixed with 95 mg of KBr and then pressed to form a tablet. The sample tablet was placed in the sample compartment and FTIR spectra were recorded in the region of 4000–400 cm 1 with 32 scans/min and resolution of 4 cm 1. FTIR spectral data pre-treatment Spectral data pre-treatment is an important step before subjecting the FTIR spectra data for multivariate analysis. It is necessary to perform this step in order to minimize the effect of light scattering, baseline variation, systematic noise, etc. [12] in all FTIR spectra of the samples. In this study, data from three different pre-treatment methods, namely standard normal variate (SNV), and first and second order derivative spectra, were compared. Chemometrics analysis Principal component analysis (PCA) and canonical variate analysis (CVA) were used to build a model for discrimination of C. longa, C. xanthorrhiza and Z. cassumunar and these analyses were performed in XLSTAT software version 2012.2.02 (Addinsoft, New York, USA).
1246
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Table 1 Sources of samples.
Table 1 (continued) Species
Species
Sample code
Sources (subdistrict, regency, province)
C. longa
CL-1 CL-2 CL-3 CL-4 CL-5 CL-6 CL-7 CL-8 CL-9 CL-10 CL-11 CL-12 CL-13 CL-14 CL-15 CL-16 CL-17 CL-18 CL-19 CL-20 CL-21 CL-22 CL-23 CL-24 CL-25 CL-26 CL-27 CL-28 CL-29
Wonogiri, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tawangmangu, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Tegalombo, Pacitan, East Java Tegalombo, Pacitan, East Java Semen, Kediri, East Java Semen, Kediri, East Java Slahung, Ponorogo, East Java Slahung, Ponorogo, East Java Ngrayun, Ponorogo, East Java Tanjungkerta, Sumedang, West Java Tanjungkerta, Sumedang, West Java Rancakalong, Sumedang, West Java Cikembar, Sukabumi, West Java Cikembar, Sukabumi, West Java Cikembar, Sukabumi, West Java Gunung Putri, Bogor, West Java Cibungbulang, Bogor, West Java Dramaga, Bogor, West Java Purworejo, Purworejo, Central Java Purworejo, Purworejo, Central Java Wonogiri, Wonogiri, Central Java Tembalang, Semarang, Central Java Kalibawang, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Pengasih, Kulonprogo, Special Region of Yogyakarta Bandar, Pacitan, East Java Tegalombo, Pacitan, East Java Tegalombo, Pacitan, East Java Ponorogo, Ponorogo, East Java
CL-30 CL-31 CL-32 CL-33 CL-34 CL-35 C. xanthorrhiza
CX-1 CX-2 CX-3 CX-4 CX-5 CX-6 CX-7 CX-8 CX-9 CX-10 CX-11 CX-12 CX-13 CX-14 CX-15 CX-16 CX-17 CX-18 CX-19 CX-20 CX-21 CX-22 CX-23 CX-24 CX-25 CX-26 CX-27 CX-28 CX-29 CX-30 CX-31
Wonogiri, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Karangpandan, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Semen, Kediri, East Java Semen, Kediri, East Java Slahung, Ponorogo, East Java Slahung, Ponorogo, East Java Ngrayun, Ponorogo, East Java Tanjungkerta, Sumedang, West Java Tanjungkerta, Sumedang, West Java Rancakalong, Sumedang, West Java Cikembar, Sukabumi, West Java Cikembar, Sukabumi, West Java Gunung Guruh, Sukabumi, West Java Gunung Putri, Bogor, West Java Ciampea, Bogor, West Java Dramaga, Bogor, West Java Pituruh, Purworejo, Central Java Purworejo, Purworejo, Central Java Purworejo, Purworejo, Central Java Wonogiri, Wonogiri, Central Java Tembalang, Semarang, Central Java Kalibawang, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Pengasih, Kulonprogo, Special Region of Yogyakarta
Z. cassumunar
Sample code
Sources (subdistrict, regency, province)
CX-32 CX-33 CX-34 CX-35
Bandar, Pacitan, East Java Bandar, Pacitan, East Java Bandar, Pacitan, East Java Ponorogo, Ponorogo, East Java
ZC-1 ZC-2 ZC-3 ZC-4 ZC-5 ZC-6 ZC-7 ZC-8 ZC-9 ZC-10 ZC-11 ZC-12 ZC-13 ZC-14 ZC-15 ZC-16 ZC-17 ZC-18 ZC-19 ZC-20 ZC-21 ZC-22 ZC-23 ZC-24 ZC-25 ZC-26
Wonogiri, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tawangmangu, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Semen, Kediri, East Java Semen, Kediri, East Java Slahung, Ponorogo, East Java Slahung, Ponorogo, East Java Tanjungkerta, Sumedang, West Java Rancakalong, Sumedang, West Java Rajapolah, Tasikmalaya, West Java Cibadak, Sukabumi, West Java Cikembar, Sukabumi, West Java Gunung Guruh, Sukabumi, West Java Gunung Putri, Bogor, West Java Ciampea, Bogor, West Java Dramaga, Bogor, West Java Purworejo, Purworejo, Central Java Purworejo, Purworejo, Central Java Wonogiri, Wonogiri, Central Java Tembalang, Semarang, Central Java Kalibawang, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Pengasih, Kulonprogo, Special Region of Yogyakarta Ponorogo, Ponorogo, East Java
ZC-27 ZC-28 ZC-29
Result and discussions Discrimination by FTIR spectral analysis Medicinal plants are complex mixture of chemicals, so their FTIR spectra will show overlapping of some characteristic absorption bands of the functional groups in the sample [11]. The FTIR spectra of C. longa, C. xanthorrhiza and Z. cassumunar are shown in Fig. 1. Under the same experimental conditions, the FTIR spectra of each species from different locations showed high similarities. As can be seen in Fig. 1, the characteristic peaks in the FTIR spectra of the three samples appeared at 3400 cm 1 corresponds to OAH absorption, at 2800–3000 cm 1 to methyl (ACH3) and methylene (ACH2) symmetric and asymmetric stretching vibration, at 1740–1680 cm 1 are attributed to C@O absorption, at 1510 cm 1 are assigned to aromatic skeletal stretching vibration, and at 1030 cm 1 are due to CAOH stretching vibration. Comparison of the FTIR spectra of C. longa and C. xanthorrhiza showed slight differences because they are from the same genus that they may have similar chemical components. However, when the FTIR spectra of C. longa and C. xanthorrhiza were compared to the FTIR spectra of Z. cassumunar obvious differences were observed. FTIR spectra could be used for the purposes of identification and discrimination of some closely-related medicinal plants. By comparing the FTIR spectra of the three species, it is recognized that there are variations in their peak positions and intensities. Z. cassumunar could be discriminated from C. longa and C. xanthorrhiza by using a peak at 1737 cm 1 for the peak only appeared in Z. cassumunar samples. Two peaks at 833 cm 1 and 816 cm 1 were only found in C. longa samples while the C. xanthorrhiza and Z. cassumunar samples
1247
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Fig. 1. Representative FTIR spectra of C. longa (A), C. xanthorrhiza (B), and Z. cassumunar (C).
Fig. 2. Representative second derivative FTIR spectra of C. longa (A), C. xanthorrhiza (B), and Z. cassumunar (C).
only showed one peak at 816 cm 1; by using this rationale, C. longa could be discriminated from the other two species in this study. Discrimination of C. xanthorrhiza from C. longa and Z. cassumunar could be observed from the intensities of OH absorption near 3400 cm 1. The intensities of the FTIR spectra of this functional group were found much greater in all C. xanthorrhiza samples than C. longa and Z. cassumunar samples. Enhancement of the spectral resolution and amplification of small differences in the ordinary spectra could be obtained using second derivative spectra. With second derivative spectra, some overlapping bands could also be resolved. Fig. 2 shows the second derivative FTIR spectra of the samples in the region 1800–400 cm 1 and it is seen that there are some differences in the spectra features. As illustrated in Fig. 2, it was clearly observed that typical positive and negative peaks at 1590 and 1579 cm 1, respectively, only appeared in C. longa sample. Discrimination of C. xanthorrhiza from the other samples could be obtained by using negative peaks at 989 cm 1 that were only observed in C. xanthorrhiza. Negative peak near 1738 cm 1 could be used for the identification of Z. cassumunar because in this sample, the peak of Z. cassumunar gave higher intensities compared to those of C. longa and C. xanthorrhiza. Beside the negative peak, discrimination of Z. cassumunar from the other two plants could be obtained by using
two positive peaks at 1452 and 1430 cm cassumunar sample.
1
that were only found in Z.
Combination of FTIR spectra and chemometrics method for discrimination of C. longa, C. xanthorrhiza and Z. cassumunar To confirm the results obtained from the visual inspection of the FTIR spectra for the discrimination of C. longa, C. xanthorrhiza and Z. cassumunar, a combination of FTIR spectra and chemometrics methods was used. Chemometrics is widely used to analyze a huge amount of data such as in FTIR spectra with an aim to resume information contained in the data matrix by reducing dimensions of the data, and finding the similarities or dissimilarities between observations and variables. In this study, some techniques in chemometrics, such as PCA and CVA, are used. Pre-treatment of FTIR spectra is a standard procedure before using the spectra in chemometrics analysis. SNV and first and second order derivative spectra were applied and compared. SNV works by calculating the standard deviation of all data points in a given FTIR spectra and then the entire FTIR spectra is normalized by this value, thus giving the FTIR spectra a unit standard deviation. SNV removes slope variation and also the scatter effects [23,24]. The first and second order derivative spectra are usually
1248
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Table 2 Eigenvalues and cumulative percentage of total variance for each PCs. Principal component
Eigenvalue
Cumulative percentage of total variance
1 2 3 4 5 6 7 8 9 10
165.536 46.443 41.023 14.145 3.911 3.262 1.760 1.066 0.470 0.415
59.309 75.949 90.647 95.714 97.116 98.285 98.915 99.297 99.465 99.614
used to eliminate the baseline drifts and for the enhancement of the small spectral features [12]. In this work, pre-treatment of the FTIR spectra using SNV gave the best result for optimum groups separation for the three species. In this case, SNV was selected to normalize the FTIR spectra before subjecting the spectra to PCA and CVA.
Principal component analysis PCA is a well-known unsupervised pattern recognition. The main objective of the PCA is to reduce the data and extract the information in order to find a combination of variables or factors for describing major trends in a data set. PCA will transform the original variables into new uncorrelated variables called principal components (PCs) that maximize the explained variance in the data on each successive component under the constraint of being orthogonal to the previous PCs [25]. In this study, PCA was employed to discriminate the samples according to the species based on the FTIR spectra in the region of 2000–400 cm 1. This region was selected because it is complex and full of information with many vibrations attributed to the chemical components in all samples. The full 99 objects 831 variables data matrices were submitted to PCA. Table 2 provides the information regarding the eigenvalues and cumulative percentage of total variance from the 10 initial PCs. About 99.61% of total variance were explained by this 10 initial PCs. Typically, PCA plot for the first two PCs is used and being the most useful in the analysis because both PCs contain the most variations in the data. The closer the PCs values, the greater the similarities among the samples. From the PCA plot, therefore, it could be obtained pattern of a sample, groupings, similarities and differences [16].
Fig. 3. PCA plot of samples: C. longa ( ), C. xanthorrhiza ( ), and Z. cassumunar (
).
Fig. 4. CVA plot of samples: C. longa ( ), C. xanthorrhiza ( ), and Z. cassumunar (
).
Fig. 3 showed the score plot derived from PCA using the first two PCs which accounted for 76% of the total variance (PC1 = 59.3% and PC2 = 16.7%). As can be seen in Fig. 3, the tested samples were clustered into three different groups. Most of the Z. cassumunar samples were separated from the C. longa and C. xanthorrhiza samples. Only three Z. cassumunar samples (ZC-3, ZC-10 and ZC-29) were detected in the C. xanthorrhiza cluster. Although C. longa and C. xanthorrhiza could be discriminated, the distance between them was so close. This situation occurred due to the fact that the chemical profile of the two plants was nearly similar as can be seen in their FTIR spectra (Fig. 1). In general, PCA could discriminate the three species.
Canonical variate analysis CVA is one of the supervised pattern recognition and widely used for multiple groups discrimination. The goal in CVA is to find linear combination of variables that exhibit maximum amonggroups variations to within-groups variations. Canonical variates (CVs) are the name for these linear combination [26]. In this work, CVA was used to discriminate the three herbs more clearly. CVA will work effectively when the number of samples is more than the number of variables. First, PCA was employed in the FTIR spectra data of samples to get PCs before building a predictive model using CVA. The criterion proposed by Kaiser was used to determine the number of PCs to be retained and used them in the CVA model. This criterion will retain PCs with eigenvalues greater than 1 because these PCs explain as much variance as are observed in the variables [27]. The CVA predictive model was built based on eight initial PCs (eigenvalues greater than 1) as input variables with the cumulative percentage of total variance greater than 99%. Separate covariance matrices were used in this classification because there are differences in the within-class covariance matrices from the Box test. From the result of CVA, the total variance from the two CVs was 100% (CV1 = 86.7% and CV2 = 13.3%). This means that it is clearly showed that 100% of the original groups were correctly classified into its own group (Fig. 4), indicating that the CVs obtained could clearly discriminate the three species. Leave-one-out cross-validation (LOOCV) method was used for evaluation of the predictive ability of this model. LOOCV works by a single training set, one of sample is removed at a time and the rest of the samples are used to build a model. Then the removed sample is treated as an unknown and its class membership is predicted [28]. As a result from the LOOCV, about 98% of all samples used in this study were correctly classified and only 1 sample (CL-4 and CL-7) was misclassified. This result indicates that the model gives satisfactory prediction of the samples tested.
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Conclusions Discrimination of C. longa, C. xanthorrhiza and Z. cassumunar was achieved by FTIR spectral analysis in combination with PCA and CVA. Through visual analysis of the ordinary and second order derivative FTIR spectra, the three species could be identified and discriminated by using their marker peaks. PCA and CVA could also discriminate the three closely-related species, in which CVA gave clearer classification based on the species. Therefore, the developed method could be applied as an effective and nondestructive method for identification and discrimination of C. longa, C. xanthorrhiza and Z. cassumunar. Acknowledgment The authors gratefully acknowledged the financial support of this research by the Fundamental Research Grant (No 45/13.24.4/ SPK/B6-PD/2009) from The Directorate of Higher Education, Ministry of Education and Culture, Republic of Indonesia. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.saa.2014.08.139. References [1] K. Tanaka, Y. Kuba, T. Sasaki, F. Hiwatashi, K. Komatsu, J. Agric. Food Chem. 56 (2008) 8787–8792. [2] S. Bua-in, Y. Paisooksantivatana, Kasetsart J. Nat. Sci. 43 (2009) 467–475. [3] M. Al-Amin, G.N.N. Sultana, C.F. Hossain, J. Ethnopharmacol. 141 (2012) 57–60. [4] L. Nahar, S.D. Sarker, Phytochemistry of the Genus Curcuma, in: P.N. Ravindran, K.N. Babu, K. Sivaraman (Eds.), Turmeric: the Genus Curcuma, CRC Press, Boca Raton, 2007, pp. 71–106.
1249
[5] S. Li, W. Yuan, G. Deng, P. Wang, P. Yang, B.B. Aggarwal, Pharm. Crops 2 (2011) 28–54. [6] A.A. Bunaciu, H.Y. Aboul-Enein, S. Fleschin, Appl. Spectrosc. Rev. 46 (2011) 251–260. [7] G.H. Lu, Q. Zhou, S.Q. Sun, K.S.Y. Leung, H. Zhang, Z.Z. Zhao, J. Mol. Struct. 883– 884 (2008) 91–98. [8] S. Sun, J. Chen, Q. Zhou, G. Lu, K. Chan, Planta Med. 76 (2010) 1987–1996. [9] H.A. Gad, S.H. El-Ahmady, M.I. Abou-Shoer, M.M. Al-Azizi, Phytochem. Anal. 24 (2013) 1–24. [10] H.B. Zou, G.S. Yang, Z.R. Qin, W.Q. Jiang, A.Q. Du, H.Y. Aboul-Enein, Anal. Lett. 38 (2005) 1457–1475. [11] H.X. Liu, Q. Zhou, S.Q. Sun, H.J. Bao, J. Mol. Struct. 883–884 (2008) 38–47. [12] Y. Chen, M.Y. Xie, Y. Yan, S.B. Zhu, S.P. Nie, C. Li, Y.X. Wang, X.F. Gong, Anal. Chim. Acta 618 (2008) 121–130. [13] P. Yang, P. Song, S.Q. Sun, Q. Zhou, S. Feng, J.X. Tao, Spectrochim. Acta A 74 (2009) 983–990. [14] L. Zhang, L. Nie, Phytochem. Anal. 21 (2010) 609–615. [15] M. Kokalj, J. Kolar, T. Trafela, S. Kreft, Phytochem. Anal. 22 (2011) 541–546. [16] J.R. Lucio-Gutierrez, J. Coello, S. Maspoch, Food Res. Int. 44 (2011) 557–565. [17] X.H. Liu, C.H. Xu, S.Q. Sun, J. Huang, K. Zhang, G.Y. Li, Y. Zhu, Q. Zhou, Z.C. Zhang, J.H. Wang, Spectrochim. Acta A 97 (2012) 290–296. [18] W. Li, Z. Cheng, Y. Wang, H. Qu, L.K. Pei, S.Q. Sun, B.L. Guo, W.H. Huang, P.G. Xiao, J. Pharm. Biomed. Anal. 72 (2013) 33–39. [19] C. Luo, T. He, Z. Chun, J. Mol. Struct. 1037 (2013) 40–48. [20] C. Xu, X. Jia, R. Xu, Y. Wang, Q. Zhou, S. Sun, Spectrochim. Acta A 114 (2013) 421–431. [21] Q. Fan, C. Chen, Y. Lin, C. Zhang, B. Liu, S. Zhao, J. Mol. Struct. 1051 (2013) 66– 71. [22] Y.K. Kwon, M.S. Ahn, J.S. Park, J.R. Liu, D.S. In, B.W. Min, S.W. Kim, J. Ginseng. Res. 38 (2014) 52–58. [23] F. Aouidi, N. Dupuy, J. Artaud, S. Roussos, M. Msallem, I. Perraud-Gaime, M. Hamdi, Food Chem. 131 (2012) 360–366. [24] A.S. Luna, A.P. da Silva, J. Ferre, R. Boque, Spectrochim. Acta A 100 (2013) 109– 114. [25] T. Rajalahti, O.M. Kvalheim, Intl. J. Pharmaceut. 417 (2011) 280–290. [26] B.K. Lavine, W.S. Rayens, Statistical Discriminant Analysis, in: S.D. Brown, R. Tauler, B. Walczak (Eds.), Comprehensive Chemometrics: Chemical and Biochemical Data Analysis, vol. 3, Elsevier, Amsterdam, 2009, pp. 517–540. [27] H.F. Kaiser, Edu. Psychol. Meas. 20 (1960) 141–151. [28] R.G. Brereton, Chemometrics: Data Analysis for the Laboratory and Chemical Plant, John Wiley and Sons, Chichester, 2003.
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
Fourier transform infrared spectroscopy combined with chemometrics for discrimination of Curcuma longa, Curcuma xanthorrhiza and Zingiber cassumunar Eti Rohaeti a, Mohamad Rafi a,b,⇑, Utami Dyah Syafitri c, Rudi Heryanto a,b a
Department of Chemistry, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Jalan Agatis Kampus IPB Dramaga, Bogor 16680, Indonesia Biopharmaca Research Center – Research and Community Empowerment Institute, Bogor Agricultural University, Jalan Taman Kencana No. 3 Kampus IPB Taman Kencana, Bogor 16128, Indonesia c Department of Statistics, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Jalan Meranti Kampus IPB Dramaga, Bogor 16680, Indonesia b
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Discrimination of C. longa, C.
xanthorrhiza and Z. cassumunar by FTIR. Principal component analysis and canonical variate analysis are used for discrimination. Canonical variate analysis gave clearer discrimination between the three species.
a r t i c l e
i n f o
Article history: Received 22 March 2014 Received in revised form 14 July 2014 Accepted 31 August 2014
Keywords: C. longa C. xanthorrhiza Z. cassumunar FTIR Chemometrics Discrimination
a b s t r a c t Turmeric (Curcuma longa), java turmeric (Curcuma xanthorrhiza) and cassumunar ginger (Zingiber cassumunar) are widely used in traditional Indonesian medicines (jamu). They have similar color for their rhizome and possess some similar uses, so it is possible to substitute one for the other. The identification and discrimination of these closely-related plants is a crucial task to ensure the quality of the raw materials. Therefore, an analytical method which is rapid, simple and accurate for discriminating these species using Fourier transform infrared spectroscopy (FTIR) combined with some chemometrics methods was developed. FTIR spectra were acquired in the mid-IR region (4000–400 cm 1). Standard normal variate, first and second order derivative spectra were compared for the spectral data. Principal component analysis (PCA) and canonical variate analysis (CVA) were used for the classification of the three species. Samples could be discriminated by visual analysis of the FTIR spectra by using their marker bands. Discrimination of the three species was also possible through the combination of the pre-processed FTIR spectra with PCA and CVA, in which CVA gave clearer discrimination. Subsequently, the developed method could be used for the identification and discrimination of the three closely-related plant species. Ó 2014 Elsevier B.V. All rights reserved.
⇑ Corresponding author at: Department of Chemistry, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Jalan Agatis Kampus IPB Dramaga, Bogor 16680, Indonesia. Tel./fax: +62 251 8624567. E-mail address: [email protected] (M. Rafi). http://dx.doi.org/10.1016/j.saa.2014.08.139 1386-1425/Ó 2014 Elsevier B.V. All rights reserved.
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Introduction Turmeric (Curcuma longa), java turmeric (Curcuma xanthorrhiza) and cassumunar ginger (Zingiber cassumunar) cultivated mostly in the Java Island are widely used in traditional Indonesian medicines (jamu). C. longa and C. xanthorrhiza are usually aromatic and carminative, and are used to treat stomachic, hepatitis, jaundice, diabetes, atherosclerosis and bacterial infections [1], while Z. cassumunar is usually used to treat various diseases such as asthma, carminative, colic, diarrhea, stomachic, muscle and joint pain [2,3]. The color of their rhizome is yellow to orange for C. longa and C. xanthorrhiza, and pale yellow to yellow for Z. cassumunar. These three plants are belongs to the Zingiberaceae family and therefore, they may have some similar chemical components. Phenolic compounds (diarylheptanoid, diarylpentanoid, phenylpropene derivative) and terpenoids (monoterpenoid, sesquiterpenoid, diterpenoid, and triterpenoid) have been identified in C. longa and some of them also present in C. xanthorrhiza. Diarylheptanoids/curcuminoids (curcumin, demethoxycurcumin, bisdemethoxycurcumin, etc.) and sesquiterpenoids (curcumene, turmerone, zingiberene, etc.) were found as a major group of compounds in C. longa and C. xanthorrhiza [4,5]. While in Z. cassumunar, phenylbutanoids ((E)-1(3,4-dimethylphenyl) butadiene), diarylheptanoids (curcumin), monoterpenoids (sabinene) and sesquiterpenoids (zingiberene) were identified [2]. Curcuminoids are responsible for the color of the three plants used in this study. Due to the similarity of the color of their rhizomes which possess some similar uses, substitution for each other could be possible, especially if in the powdered form and if the price of C. xanthorrhiza is much higher than C. longa or Z. cassumunar. If the substitution occured, it could be a serious problem because of inconsistency in their biological properties. In this case, the identification and discrimination of these closely-related plants are crucial in order to ensure the quality, safety and efficacy of the raw materials before they are converted to final products. Therefore, a rapid, simple and accurate analytical method is essentially required for the discrimination of these species. Several analytical techniques such as spectroscopy (UV–Vis, FTIR, NMR, and MS) and chromatography (TLC, HPLC, and GC) have been used in the development of method for identification and discrimination of medicinal plants. Among these techniques, Fourier transform infrared (FTIR) spectroscopy may be an attractive option because it can meet the criteria of efficient analysis, i.e. easy to use, fast, and inexpensive [6]. Although medicinal plants will contain a large variety of chemical components, their FTIR spectra sometimes are found to differ even in the same species [7]. FTIR has been widely used and is a well-established tool for quality control in various industries, including herbal industry [8]. FTIR spectra contain complex data information describing the overall chemical signal in a sample. Changes in the position and intensity of bands in the FTIR spectra would be associated with the changes in the chemical composition of a sample. Therefore, FTIR spectra could be used to discriminate closely-related species, although the composition of a chemical compound is certainly not known [8]. Obvious advantages of FTIR application to discriminate different medicinal plants are not only effective and specific, but also rapid and non-separative. Discrimination by visual inspection in the FTIR spectra is not easy because the FTIR spectra pattern is very complex, and to address this concern, the aid from chemometrics methods is needed [9]. The advantage of using chemometrics for the interpretation of FTIR is the ability to link the spectral pattern with hidden information contained in a sample [10]. FTIR spectral analysis alone or in combination with chemometrics methods has been extensively used for species identification and discrimination of some closely-related species [11–22]. Because of the advantages
1245
of FTIR spectroscopy, we used this techniques in combination with chemometrics methods for the first time to develop a qualitative method for discrimination of C. Longa, C. xanthorrhiza and Z. Cassumunar. This combined method was successfully applied for the identification and discrimination of the three species.
Material and methods Chemicals Potassium bromide (KBr) for spectroscopy was purchased from Sigma–Aldrich (St Louis, USA). Analytical-grade ethanol for solvent extraction was obtained from Merck (Darmstadt, Germany). Sample preparation Ninety-nine samples consisting of 35 samples of C. longa, 35 samples of C. xanthorrhiza and 29 samples of Z. cassumunar were collected during 2008–2010 from 11 regencies (2–5 samples from each regency) in Java Island, Indonesia: Bogor, Sukabumi, Sumedang, Purworejo, Wonogiri, Karanganyar, Semarang, Kulonprogo, Pacitan, Ponorogo, and Kediri (Table 1). Voucher specimens were deposited at Biopharmaca Research Center, Bogor Agricultural University, Indonesia. All samples were sieved, dried, and pulverized prior to use. There are several methods to obtained IR spectra for the purpose of identification and discrimination as described by Zou et al. [10]. In this experiment, the samples were extracted with a solvent and after evaporating the solvent, the extracts were blended with KBr. This method will provide better resolution in the identification and discrimination of closely related plants [10]. Ethanol was used to extract the samples and after the extraction process, the ethanol was evaporated by rotary evaporator. These ethanol extracts were then used for FTIR measurement. FTIR spectroscopy measurement FTIR spectra were obtained using a Bruker Tensor 37 FTIR spectrophotometer equipped with deuterated triglycine-sulphate (DTGS) as detector and controlled by OPUS 4.2 software (Bruker, Germany). About 5 mg of ethanol extract of each sample was mixed with 95 mg of KBr and then pressed to form a tablet. The sample tablet was placed in the sample compartment and FTIR spectra were recorded in the region of 4000–400 cm 1 with 32 scans/min and resolution of 4 cm 1. FTIR spectral data pre-treatment Spectral data pre-treatment is an important step before subjecting the FTIR spectra data for multivariate analysis. It is necessary to perform this step in order to minimize the effect of light scattering, baseline variation, systematic noise, etc. [12] in all FTIR spectra of the samples. In this study, data from three different pre-treatment methods, namely standard normal variate (SNV), and first and second order derivative spectra, were compared. Chemometrics analysis Principal component analysis (PCA) and canonical variate analysis (CVA) were used to build a model for discrimination of C. longa, C. xanthorrhiza and Z. cassumunar and these analyses were performed in XLSTAT software version 2012.2.02 (Addinsoft, New York, USA).
1246
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Table 1 Sources of samples.
Table 1 (continued) Species
Species
Sample code
Sources (subdistrict, regency, province)
C. longa
CL-1 CL-2 CL-3 CL-4 CL-5 CL-6 CL-7 CL-8 CL-9 CL-10 CL-11 CL-12 CL-13 CL-14 CL-15 CL-16 CL-17 CL-18 CL-19 CL-20 CL-21 CL-22 CL-23 CL-24 CL-25 CL-26 CL-27 CL-28 CL-29
Wonogiri, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tawangmangu, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Tegalombo, Pacitan, East Java Tegalombo, Pacitan, East Java Semen, Kediri, East Java Semen, Kediri, East Java Slahung, Ponorogo, East Java Slahung, Ponorogo, East Java Ngrayun, Ponorogo, East Java Tanjungkerta, Sumedang, West Java Tanjungkerta, Sumedang, West Java Rancakalong, Sumedang, West Java Cikembar, Sukabumi, West Java Cikembar, Sukabumi, West Java Cikembar, Sukabumi, West Java Gunung Putri, Bogor, West Java Cibungbulang, Bogor, West Java Dramaga, Bogor, West Java Purworejo, Purworejo, Central Java Purworejo, Purworejo, Central Java Wonogiri, Wonogiri, Central Java Tembalang, Semarang, Central Java Kalibawang, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Pengasih, Kulonprogo, Special Region of Yogyakarta Bandar, Pacitan, East Java Tegalombo, Pacitan, East Java Tegalombo, Pacitan, East Java Ponorogo, Ponorogo, East Java
CL-30 CL-31 CL-32 CL-33 CL-34 CL-35 C. xanthorrhiza
CX-1 CX-2 CX-3 CX-4 CX-5 CX-6 CX-7 CX-8 CX-9 CX-10 CX-11 CX-12 CX-13 CX-14 CX-15 CX-16 CX-17 CX-18 CX-19 CX-20 CX-21 CX-22 CX-23 CX-24 CX-25 CX-26 CX-27 CX-28 CX-29 CX-30 CX-31
Wonogiri, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Karangpandan, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Semen, Kediri, East Java Semen, Kediri, East Java Slahung, Ponorogo, East Java Slahung, Ponorogo, East Java Ngrayun, Ponorogo, East Java Tanjungkerta, Sumedang, West Java Tanjungkerta, Sumedang, West Java Rancakalong, Sumedang, West Java Cikembar, Sukabumi, West Java Cikembar, Sukabumi, West Java Gunung Guruh, Sukabumi, West Java Gunung Putri, Bogor, West Java Ciampea, Bogor, West Java Dramaga, Bogor, West Java Pituruh, Purworejo, Central Java Purworejo, Purworejo, Central Java Purworejo, Purworejo, Central Java Wonogiri, Wonogiri, Central Java Tembalang, Semarang, Central Java Kalibawang, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Pengasih, Kulonprogo, Special Region of Yogyakarta
Z. cassumunar
Sample code
Sources (subdistrict, regency, province)
CX-32 CX-33 CX-34 CX-35
Bandar, Pacitan, East Java Bandar, Pacitan, East Java Bandar, Pacitan, East Java Ponorogo, Ponorogo, East Java
ZC-1 ZC-2 ZC-3 ZC-4 ZC-5 ZC-6 ZC-7 ZC-8 ZC-9 ZC-10 ZC-11 ZC-12 ZC-13 ZC-14 ZC-15 ZC-16 ZC-17 ZC-18 ZC-19 ZC-20 ZC-21 ZC-22 ZC-23 ZC-24 ZC-25 ZC-26
Wonogiri, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Ngadirejo, Wonogiri, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tembalang, Semarang, Central Java Tawangmangu, Karanganyar, Central Java Tawangmangu, Karanganyar, Central Java Semen, Kediri, East Java Semen, Kediri, East Java Slahung, Ponorogo, East Java Slahung, Ponorogo, East Java Tanjungkerta, Sumedang, West Java Rancakalong, Sumedang, West Java Rajapolah, Tasikmalaya, West Java Cibadak, Sukabumi, West Java Cikembar, Sukabumi, West Java Gunung Guruh, Sukabumi, West Java Gunung Putri, Bogor, West Java Ciampea, Bogor, West Java Dramaga, Bogor, West Java Purworejo, Purworejo, Central Java Purworejo, Purworejo, Central Java Wonogiri, Wonogiri, Central Java Tembalang, Semarang, Central Java Kalibawang, Kulonprogo, Special Region of Yogyakarta Wates, Kulonprogo, Special Region of Yogyakarta Pengasih, Kulonprogo, Special Region of Yogyakarta Ponorogo, Ponorogo, East Java
ZC-27 ZC-28 ZC-29
Result and discussions Discrimination by FTIR spectral analysis Medicinal plants are complex mixture of chemicals, so their FTIR spectra will show overlapping of some characteristic absorption bands of the functional groups in the sample [11]. The FTIR spectra of C. longa, C. xanthorrhiza and Z. cassumunar are shown in Fig. 1. Under the same experimental conditions, the FTIR spectra of each species from different locations showed high similarities. As can be seen in Fig. 1, the characteristic peaks in the FTIR spectra of the three samples appeared at 3400 cm 1 corresponds to OAH absorption, at 2800–3000 cm 1 to methyl (ACH3) and methylene (ACH2) symmetric and asymmetric stretching vibration, at 1740–1680 cm 1 are attributed to C@O absorption, at 1510 cm 1 are assigned to aromatic skeletal stretching vibration, and at 1030 cm 1 are due to CAOH stretching vibration. Comparison of the FTIR spectra of C. longa and C. xanthorrhiza showed slight differences because they are from the same genus that they may have similar chemical components. However, when the FTIR spectra of C. longa and C. xanthorrhiza were compared to the FTIR spectra of Z. cassumunar obvious differences were observed. FTIR spectra could be used for the purposes of identification and discrimination of some closely-related medicinal plants. By comparing the FTIR spectra of the three species, it is recognized that there are variations in their peak positions and intensities. Z. cassumunar could be discriminated from C. longa and C. xanthorrhiza by using a peak at 1737 cm 1 for the peak only appeared in Z. cassumunar samples. Two peaks at 833 cm 1 and 816 cm 1 were only found in C. longa samples while the C. xanthorrhiza and Z. cassumunar samples
1247
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Fig. 1. Representative FTIR spectra of C. longa (A), C. xanthorrhiza (B), and Z. cassumunar (C).
Fig. 2. Representative second derivative FTIR spectra of C. longa (A), C. xanthorrhiza (B), and Z. cassumunar (C).
only showed one peak at 816 cm 1; by using this rationale, C. longa could be discriminated from the other two species in this study. Discrimination of C. xanthorrhiza from C. longa and Z. cassumunar could be observed from the intensities of OH absorption near 3400 cm 1. The intensities of the FTIR spectra of this functional group were found much greater in all C. xanthorrhiza samples than C. longa and Z. cassumunar samples. Enhancement of the spectral resolution and amplification of small differences in the ordinary spectra could be obtained using second derivative spectra. With second derivative spectra, some overlapping bands could also be resolved. Fig. 2 shows the second derivative FTIR spectra of the samples in the region 1800–400 cm 1 and it is seen that there are some differences in the spectra features. As illustrated in Fig. 2, it was clearly observed that typical positive and negative peaks at 1590 and 1579 cm 1, respectively, only appeared in C. longa sample. Discrimination of C. xanthorrhiza from the other samples could be obtained by using negative peaks at 989 cm 1 that were only observed in C. xanthorrhiza. Negative peak near 1738 cm 1 could be used for the identification of Z. cassumunar because in this sample, the peak of Z. cassumunar gave higher intensities compared to those of C. longa and C. xanthorrhiza. Beside the negative peak, discrimination of Z. cassumunar from the other two plants could be obtained by using
two positive peaks at 1452 and 1430 cm cassumunar sample.
1
that were only found in Z.
Combination of FTIR spectra and chemometrics method for discrimination of C. longa, C. xanthorrhiza and Z. cassumunar To confirm the results obtained from the visual inspection of the FTIR spectra for the discrimination of C. longa, C. xanthorrhiza and Z. cassumunar, a combination of FTIR spectra and chemometrics methods was used. Chemometrics is widely used to analyze a huge amount of data such as in FTIR spectra with an aim to resume information contained in the data matrix by reducing dimensions of the data, and finding the similarities or dissimilarities between observations and variables. In this study, some techniques in chemometrics, such as PCA and CVA, are used. Pre-treatment of FTIR spectra is a standard procedure before using the spectra in chemometrics analysis. SNV and first and second order derivative spectra were applied and compared. SNV works by calculating the standard deviation of all data points in a given FTIR spectra and then the entire FTIR spectra is normalized by this value, thus giving the FTIR spectra a unit standard deviation. SNV removes slope variation and also the scatter effects [23,24]. The first and second order derivative spectra are usually
1248
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Table 2 Eigenvalues and cumulative percentage of total variance for each PCs. Principal component
Eigenvalue
Cumulative percentage of total variance
1 2 3 4 5 6 7 8 9 10
165.536 46.443 41.023 14.145 3.911 3.262 1.760 1.066 0.470 0.415
59.309 75.949 90.647 95.714 97.116 98.285 98.915 99.297 99.465 99.614
used to eliminate the baseline drifts and for the enhancement of the small spectral features [12]. In this work, pre-treatment of the FTIR spectra using SNV gave the best result for optimum groups separation for the three species. In this case, SNV was selected to normalize the FTIR spectra before subjecting the spectra to PCA and CVA.
Principal component analysis PCA is a well-known unsupervised pattern recognition. The main objective of the PCA is to reduce the data and extract the information in order to find a combination of variables or factors for describing major trends in a data set. PCA will transform the original variables into new uncorrelated variables called principal components (PCs) that maximize the explained variance in the data on each successive component under the constraint of being orthogonal to the previous PCs [25]. In this study, PCA was employed to discriminate the samples according to the species based on the FTIR spectra in the region of 2000–400 cm 1. This region was selected because it is complex and full of information with many vibrations attributed to the chemical components in all samples. The full 99 objects 831 variables data matrices were submitted to PCA. Table 2 provides the information regarding the eigenvalues and cumulative percentage of total variance from the 10 initial PCs. About 99.61% of total variance were explained by this 10 initial PCs. Typically, PCA plot for the first two PCs is used and being the most useful in the analysis because both PCs contain the most variations in the data. The closer the PCs values, the greater the similarities among the samples. From the PCA plot, therefore, it could be obtained pattern of a sample, groupings, similarities and differences [16].
Fig. 3. PCA plot of samples: C. longa ( ), C. xanthorrhiza ( ), and Z. cassumunar (
).
Fig. 4. CVA plot of samples: C. longa ( ), C. xanthorrhiza ( ), and Z. cassumunar (
).
Fig. 3 showed the score plot derived from PCA using the first two PCs which accounted for 76% of the total variance (PC1 = 59.3% and PC2 = 16.7%). As can be seen in Fig. 3, the tested samples were clustered into three different groups. Most of the Z. cassumunar samples were separated from the C. longa and C. xanthorrhiza samples. Only three Z. cassumunar samples (ZC-3, ZC-10 and ZC-29) were detected in the C. xanthorrhiza cluster. Although C. longa and C. xanthorrhiza could be discriminated, the distance between them was so close. This situation occurred due to the fact that the chemical profile of the two plants was nearly similar as can be seen in their FTIR spectra (Fig. 1). In general, PCA could discriminate the three species.
Canonical variate analysis CVA is one of the supervised pattern recognition and widely used for multiple groups discrimination. The goal in CVA is to find linear combination of variables that exhibit maximum amonggroups variations to within-groups variations. Canonical variates (CVs) are the name for these linear combination [26]. In this work, CVA was used to discriminate the three herbs more clearly. CVA will work effectively when the number of samples is more than the number of variables. First, PCA was employed in the FTIR spectra data of samples to get PCs before building a predictive model using CVA. The criterion proposed by Kaiser was used to determine the number of PCs to be retained and used them in the CVA model. This criterion will retain PCs with eigenvalues greater than 1 because these PCs explain as much variance as are observed in the variables [27]. The CVA predictive model was built based on eight initial PCs (eigenvalues greater than 1) as input variables with the cumulative percentage of total variance greater than 99%. Separate covariance matrices were used in this classification because there are differences in the within-class covariance matrices from the Box test. From the result of CVA, the total variance from the two CVs was 100% (CV1 = 86.7% and CV2 = 13.3%). This means that it is clearly showed that 100% of the original groups were correctly classified into its own group (Fig. 4), indicating that the CVs obtained could clearly discriminate the three species. Leave-one-out cross-validation (LOOCV) method was used for evaluation of the predictive ability of this model. LOOCV works by a single training set, one of sample is removed at a time and the rest of the samples are used to build a model. Then the removed sample is treated as an unknown and its class membership is predicted [28]. As a result from the LOOCV, about 98% of all samples used in this study were correctly classified and only 1 sample (CL-4 and CL-7) was misclassified. This result indicates that the model gives satisfactory prediction of the samples tested.
E. Rohaeti et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1244–1249
Conclusions Discrimination of C. longa, C. xanthorrhiza and Z. cassumunar was achieved by FTIR spectral analysis in combination with PCA and CVA. Through visual analysis of the ordinary and second order derivative FTIR spectra, the three species could be identified and discriminated by using their marker peaks. PCA and CVA could also discriminate the three closely-related species, in which CVA gave clearer classification based on the species. Therefore, the developed method could be applied as an effective and nondestructive method for identification and discrimination of C. longa, C. xanthorrhiza and Z. cassumunar. Acknowledgment The authors gratefully acknowledged the financial support of this research by the Fundamental Research Grant (No 45/13.24.4/ SPK/B6-PD/2009) from The Directorate of Higher Education, Ministry of Education and Culture, Republic of Indonesia. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.saa.2014.08.139. References [1] K. Tanaka, Y. Kuba, T. Sasaki, F. Hiwatashi, K. Komatsu, J. Agric. Food Chem. 56 (2008) 8787–8792. [2] S. Bua-in, Y. Paisooksantivatana, Kasetsart J. Nat. Sci. 43 (2009) 467–475. [3] M. Al-Amin, G.N.N. Sultana, C.F. Hossain, J. Ethnopharmacol. 141 (2012) 57–60. [4] L. Nahar, S.D. Sarker, Phytochemistry of the Genus Curcuma, in: P.N. Ravindran, K.N. Babu, K. Sivaraman (Eds.), Turmeric: the Genus Curcuma, CRC Press, Boca Raton, 2007, pp. 71–106.
1249
[5] S. Li, W. Yuan, G. Deng, P. Wang, P. Yang, B.B. Aggarwal, Pharm. Crops 2 (2011) 28–54. [6] A.A. Bunaciu, H.Y. Aboul-Enein, S. Fleschin, Appl. Spectrosc. Rev. 46 (2011) 251–260. [7] G.H. Lu, Q. Zhou, S.Q. Sun, K.S.Y. Leung, H. Zhang, Z.Z. Zhao, J. Mol. Struct. 883– 884 (2008) 91–98. [8] S. Sun, J. Chen, Q. Zhou, G. Lu, K. Chan, Planta Med. 76 (2010) 1987–1996. [9] H.A. Gad, S.H. El-Ahmady, M.I. Abou-Shoer, M.M. Al-Azizi, Phytochem. Anal. 24 (2013) 1–24. [10] H.B. Zou, G.S. Yang, Z.R. Qin, W.Q. Jiang, A.Q. Du, H.Y. Aboul-Enein, Anal. Lett. 38 (2005) 1457–1475. [11] H.X. Liu, Q. Zhou, S.Q. Sun, H.J. Bao, J. Mol. Struct. 883–884 (2008) 38–47. [12] Y. Chen, M.Y. Xie, Y. Yan, S.B. Zhu, S.P. Nie, C. Li, Y.X. Wang, X.F. Gong, Anal. Chim. Acta 618 (2008) 121–130. [13] P. Yang, P. Song, S.Q. Sun, Q. Zhou, S. Feng, J.X. Tao, Spectrochim. Acta A 74 (2009) 983–990. [14] L. Zhang, L. Nie, Phytochem. Anal. 21 (2010) 609–615. [15] M. Kokalj, J. Kolar, T. Trafela, S. Kreft, Phytochem. Anal. 22 (2011) 541–546. [16] J.R. Lucio-Gutierrez, J. Coello, S. Maspoch, Food Res. Int. 44 (2011) 557–565. [17] X.H. Liu, C.H. Xu, S.Q. Sun, J. Huang, K. Zhang, G.Y. Li, Y. Zhu, Q. Zhou, Z.C. Zhang, J.H. Wang, Spectrochim. Acta A 97 (2012) 290–296. [18] W. Li, Z. Cheng, Y. Wang, H. Qu, L.K. Pei, S.Q. Sun, B.L. Guo, W.H. Huang, P.G. Xiao, J. Pharm. Biomed. Anal. 72 (2013) 33–39. [19] C. Luo, T. He, Z. Chun, J. Mol. Struct. 1037 (2013) 40–48. [20] C. Xu, X. Jia, R. Xu, Y. Wang, Q. Zhou, S. Sun, Spectrochim. Acta A 114 (2013) 421–431. [21] Q. Fan, C. Chen, Y. Lin, C. Zhang, B. Liu, S. Zhao, J. Mol. Struct. 1051 (2013) 66– 71. [22] Y.K. Kwon, M.S. Ahn, J.S. Park, J.R. Liu, D.S. In, B.W. Min, S.W. Kim, J. Ginseng. Res. 38 (2014) 52–58. [23] F. Aouidi, N. Dupuy, J. Artaud, S. Roussos, M. Msallem, I. Perraud-Gaime, M. Hamdi, Food Chem. 131 (2012) 360–366. [24] A.S. Luna, A.P. da Silva, J. Ferre, R. Boque, Spectrochim. Acta A 100 (2013) 109– 114. [25] T. Rajalahti, O.M. Kvalheim, Intl. J. Pharmaceut. 417 (2011) 280–290. [26] B.K. Lavine, W.S. Rayens, Statistical Discriminant Analysis, in: S.D. Brown, R. Tauler, B. Walczak (Eds.), Comprehensive Chemometrics: Chemical and Biochemical Data Analysis, vol. 3, Elsevier, Amsterdam, 2009, pp. 517–540. [27] H.F. Kaiser, Edu. Psychol. Meas. 20 (1960) 141–151. [28] R.G. Brereton, Chemometrics: Data Analysis for the Laboratory and Chemical Plant, John Wiley and Sons, Chichester, 2003.
