+ All Categories
Home > Documents > Differentiation of Anatolian honey samples from different botanical ...

Differentiation of Anatolian honey samples from different botanical ...

Date post: 02-Jan-2017
Category:
Upload: hakhanh
View: 222 times
Download: 0 times
Share this document with a friend
7
Differentiation of Anatolian honey samples from different botanical origins by ATR-FTIR spectroscopy using multivariate analysis Seher Gok a , Mete Severcan b , Erik Goormaghtigh c , Irfan Kandemir d , Feride Severcan a,a Department of Biological Sciences, Middle East Technical University, 06531 Ankara, Turkey b Department of Electrical and Electronic Engineering, Middle East Technical University, 06531 Ankara, Turkey c Center for Structural Biology and Bioinformatics, Laboratory for the Structure and Function of Biological Membranes, Université Libre de Bruxelles, Brussels, Belgium d Department of Biology, Ankara University, Ankara, Turkey article info Article history: Received 27 March 2014 Received in revised form 8 August 2014 Accepted 10 August 2014 Available online 20 August 2014 Keywords: Honey Botanical origin ATR-FTIR spectroscopy Multivariate analysis Hierarchical Cluster Analysis Principal Component Analysis abstract Botanical origin of the nectar predominantly affects the chemical composition of honey. Analytical techniques used for reliable honey authentication are mostly time consuming and expensive. Addition- ally, they cannot provide 100% efficiency in accurate authentication. Therefore, alternatives for the determination of floral origin of honey need to be developed. This study aims to discriminate character- istic Anatolian honey samples from different botanical origins based on the differences in their molecular content, rather than giving numerical information about the constituents of samples. Another scope of the study is to differentiate inauthentic honey samples from the natural ones precisely. All samples were tested via unsupervised pattern recognition procedures like hierarchical clustering and Principal Compo- nent Analysis (PCA). Discrimination of sample groups was achieved successfully with hierarchical cluster- ing over the spectral range of 1800–750 cm 1 which suggests a good predictive capability of Fourier Transform Infrared (FTIR) spectroscopy and chemometry for the determination of honey floral source. Ó 2014 Elsevier Ltd. All rights reserved. 1. Introduction Honey characteristics depend primarily on the botanical origin of nectar. Floral source of the nectar predominantly affects the chemical composition of honey in terms of its protein, carbohy- drate, enzyme, mineral and organic acid content. It is known that there are more than 100 different botanical origins for the honey. According to the Codex Alimentarius Standard for Honey and the European Union Council Directive related to honey; ‘‘The use of a botanical designation of honey is allowed if it originates predomi- nantly from the indicated floral source’’. Botanical denomination is used for the presentation of more than 50% of honey products. Particularly, unifloral honey has a higher demand and commercial value in the market. However, 60% of indications related to floral origin made by beekeepers are incorrect. Therefore reliable authentication by analytical techniques is important for certifica- tion and quality control of honey (Bryant & Jones, 2001). In the European Union the composition, manufacture and marketing of honey is regulated by the Community Directive 74/409/EEC. As a community standard, information referring to honey’s geographi- cal and floral origin must be supplemented (Radovic, Goodacre, & Anklam, 2001). Authentication of honey has primary importance for both industries and consumers. Demand for ‘‘natural’’ honey has been increasing especially in medical market due to its thera- peutic effects. In addition, from the economic perspective, authen- tication is needed to avoid unfair competition which can lead to a destabilization in market (Cordella, Moussa, Martel, Sbirrazzuoli, & Lizzani-Cuvelier, 2002). Turkey has suitable geographical and cli- matic conditions for apiculture where approximately 6,600,000 hives resided and lead to a production of 94,694 tones of honey in the year 2013 (TUIK, 2013) and one of the most important honey producer and exporter in the worldwide. Anatolian honeys are rich in pollen types per sample and 85% of the world’s floral types can be found in Turkish honeys. Despite the great diversity of honeys produced in Anatolia, there have been limited studies for the char- acterisation and classification by botanical or geographical origin. Also these publications are limited to compositional analysis (Kayacier & Karaman, 2008; Kucuk et al., 2007; Senyuva et al., 2009; Silici & Gokceoglu, 2007). With this study, for the first time we have classified the wide range of different authenticated floral types of honey from Anatolia using spectroscopic and chemometric methods. Many analytical techniques have been applied for reliable authenticity testing of honey like high performance liquid chromatography (HPLC) (Swallow & Low, 1990), nuclear magnetic http://dx.doi.org/10.1016/j.foodchem.2014.08.040 0308-8146/Ó 2014 Elsevier Ltd. All rights reserved. Corresponding author. Tel.: +90 0312 210 5157; fax: +90 0312 210 7970. E-mail address: [email protected] (F. Severcan). Food Chemistry 170 (2015) 234–240 Contents lists available at ScienceDirect Food Chemistry journal homepage: www.elsevier.com/locate/foodchem
Transcript
Page 1: Differentiation of Anatolian honey samples from different botanical ...

Food Chemistry 170 (2015) 234–240

Contents lists available at ScienceDirect

Food Chemistry

journal homepage: www.elsevier .com/locate / foodchem

Differentiation of Anatolian honey samples from different botanicalorigins by ATR-FTIR spectroscopy using multivariate analysis

http://dx.doi.org/10.1016/j.foodchem.2014.08.0400308-8146/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +90 0312 210 5157; fax: +90 0312 210 7970.E-mail address: [email protected] (F. Severcan).

Seher Gok a, Mete Severcan b, Erik Goormaghtigh c, Irfan Kandemir d, Feride Severcan a,⇑a Department of Biological Sciences, Middle East Technical University, 06531 Ankara, Turkeyb Department of Electrical and Electronic Engineering, Middle East Technical University, 06531 Ankara, Turkeyc Center for Structural Biology and Bioinformatics, Laboratory for the Structure and Function of Biological Membranes, Université Libre de Bruxelles, Brussels, Belgiumd Department of Biology, Ankara University, Ankara, Turkey

a r t i c l e i n f o a b s t r a c t

Article history:Received 27 March 2014Received in revised form 8 August 2014Accepted 10 August 2014Available online 20 August 2014

Keywords:HoneyBotanical originATR-FTIR spectroscopyMultivariate analysisHierarchical Cluster AnalysisPrincipal Component Analysis

Botanical origin of the nectar predominantly affects the chemical composition of honey. Analyticaltechniques used for reliable honey authentication are mostly time consuming and expensive. Addition-ally, they cannot provide 100% efficiency in accurate authentication. Therefore, alternatives for thedetermination of floral origin of honey need to be developed. This study aims to discriminate character-istic Anatolian honey samples from different botanical origins based on the differences in their molecularcontent, rather than giving numerical information about the constituents of samples. Another scope ofthe study is to differentiate inauthentic honey samples from the natural ones precisely. All samples weretested via unsupervised pattern recognition procedures like hierarchical clustering and Principal Compo-nent Analysis (PCA). Discrimination of sample groups was achieved successfully with hierarchical cluster-ing over the spectral range of 1800–750 cm�1 which suggests a good predictive capability of FourierTransform Infrared (FTIR) spectroscopy and chemometry for the determination of honey floral source.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Honey characteristics depend primarily on the botanical originof nectar. Floral source of the nectar predominantly affects thechemical composition of honey in terms of its protein, carbohy-drate, enzyme, mineral and organic acid content. It is known thatthere are more than 100 different botanical origins for the honey.According to the Codex Alimentarius Standard for Honey and theEuropean Union Council Directive related to honey; ‘‘The use of abotanical designation of honey is allowed if it originates predomi-nantly from the indicated floral source’’. Botanical denomination isused for the presentation of more than 50% of honey products.Particularly, unifloral honey has a higher demand and commercialvalue in the market. However, 60% of indications related to floralorigin made by beekeepers are incorrect. Therefore reliableauthentication by analytical techniques is important for certifica-tion and quality control of honey (Bryant & Jones, 2001). In theEuropean Union the composition, manufacture and marketing ofhoney is regulated by the Community Directive 74/409/EEC. As acommunity standard, information referring to honey’s geographi-cal and floral origin must be supplemented (Radovic, Goodacre, &

Anklam, 2001). Authentication of honey has primary importancefor both industries and consumers. Demand for ‘‘natural’’ honeyhas been increasing especially in medical market due to its thera-peutic effects. In addition, from the economic perspective, authen-tication is needed to avoid unfair competition which can lead to adestabilization in market (Cordella, Moussa, Martel, Sbirrazzuoli, &Lizzani-Cuvelier, 2002). Turkey has suitable geographical and cli-matic conditions for apiculture where approximately 6,600,000hives resided and lead to a production of 94,694 tones of honeyin the year 2013 (TUIK, 2013) and one of the most important honeyproducer and exporter in the worldwide. Anatolian honeys are richin pollen types per sample and 85% of the world’s floral types canbe found in Turkish honeys. Despite the great diversity of honeysproduced in Anatolia, there have been limited studies for the char-acterisation and classification by botanical or geographical origin.Also these publications are limited to compositional analysis(Kayacier & Karaman, 2008; Kucuk et al., 2007; Senyuva et al.,2009; Silici & Gokceoglu, 2007). With this study, for the first timewe have classified the wide range of different authenticated floraltypes of honey from Anatolia using spectroscopic and chemometricmethods.

Many analytical techniques have been applied for reliableauthenticity testing of honey like high performance liquidchromatography (HPLC) (Swallow & Low, 1990), nuclear magnetic

Page 2: Differentiation of Anatolian honey samples from different botanical ...

S. Gok et al. / Food Chemistry 170 (2015) 234–240 235

resonance (NMR) (Lindner, Bermann, & Gamarnik, 1996), gas chro-matography (Low & South, 1995) and carbon isotope ratio analysis(White, Winters, Martin, & Rossmann, 1998). These techniquesused for reliable honey authentication are mostly time consumingand expensive. Additionally, they cannot provide 100% success inauthentication. Infrared (IR) spectroscopy has been preferred as arapid, non-destructive, reagent-free, operator independent andcheap technique in food industry for the quantification of variousfood samples (Chalmers & Griffiths, 2002; Li-Chan, Chalmers, &Griffiths, 2010). IR spectroscopy was applied in different honeysamples for the determination of botanical or geographical origin,detection of adulteration and for the quantification of fructose, glu-cose, sucrose, maltose, pH value and electrical conductivity (Chung,Ku, & Lee, 1999; Lichtenberg-Kraag, Hedtke, & Bienefeld, 2002;Ruoff, 2006; Tewari & Irudayaraj, 2004). Chemometric methodsbased on Fourier Transform Infrared (FTIR) spectroscopy were alsoused for honey adulteration (Rios-Corripio, Rojas-López, &Delgado-Macuil, 2012; Subari, Saleh, Shakaff, & Zakaria, 2012)and characterisation with limited number of Mexican honeys(Rios-Corripio, Rios-Leal, Rojas-López, & Delgado-Macuil, 2011).Etzold and Lichtenberg-Kraag (2008) have developed FTIR basedPCA calibration models with German honeys. Although, there havebeen many attempts for searching alternative methods for authen-tication of honey, some of these studies have been limited withcertain number of unifloral honey sources and have not beentested sufficiently with polyfloral samples (Ruoff, 2006) and withAnatolian honeys specifically.

In the current study, it was aimed to estimate botanical origin ofhoney samples that are specific to Anatolia by applying two differ-ent multivariate analysis techniques to the Attenuated TotalReflectance (ATR)-FTIR spectroscopic data. With this work, it wasintended to exemplify the usage and success of ATR-FTIR spectros-copy coupled with multivariate analysis in botanical origin assign-ment with a high number of sample groups. This work will alsoprovide basis to honey adulteration determination studies.

2. Experimental

2.1. Samples

A total of 144 honey samples were collected from different geo-graphical regions of Turkey. The majority of samples used in thisstudy were procured from well known certificated honey brandswhich have BRC (British Retail Consortium) certificate andofficially declare that honeys are subjected to all chemical andphysical analysis to detect quality and purity in addition todescriptive organoleptic and microscopic analysis to determinefloral and regional origins. Some samples were collected directlyfrom the primary producers. The region and origin of productionwere known for all samples.

Flower originated (polyfloral (n = 30), anzer (n = 3), organic(n = 13), Taurus flower (n = 6)), tree originated (pine (n = 22),chestnut (n = 10), cedar (n = 6)) and rhododendron honey (n = 30)samples in addition to fake (adulterated) honey (n = 6), maplesyrup (n = 6), fructose syrup (n = 6) and grape molasses (n = 6)samples were included in the study. The sample size of each groupis indicated as ‘‘n’’.

Honey samples were grouped as tree and flower originated onesbasically. Flower originated group is composed of polyfloral hon-eys, collected from different regions of Turkey, which are Anzerhoney, organic flower honey and Taurus flower honey. Anzerhoney, composed of nectar, mainly collected from Anzer plant(thymus species) in the narrow region located in Rize/Ikizdere/Anzer. It has been largely studied in terms of its medicinal proper-ties that cause its market price to be 10 times higher than otherhoneys. Organic honeys are collected from the well known brands

which have ‘‘organic’’ certificate. Taurus flower honey is collectedfrom the Taurus Mountains located at the south Mediterraneanregion of Anatolia.

Rhododendron honey, locally called as ‘‘mad honey’’ or ‘‘toxichoney’’, is made up of spring flowers of Rhododendron ponticum(rhododendron plant). Rhododendrons mainly grow in the easternBlack Sea Region of Turkey. Their phenolic content and antimicro-bial activities are quite different from the other honey plantspecies. Nectar contains andromedotoxin, which causes variousphysiological effects in humans (Onat, Yegen, Lawrence, Oktay, &Oktay, 1991).

Chestnut honey is produced from both nectar and secretumcollection by honey bee. These are collected from various regionsof Anatolia.

Pine honey is a kind of honeydew honey. It is produced viausing the secretum of an insect (Marchalina hellenica) living inthe trunk of pine tree and collected by bees. Pine honey is a specificendemic product, and can be found only in Turkey and Greece.

Cedar honey, used in this study, was collected from the TaurusMountains in the Mediterranean region of Turkey, and is mainlyoriginated from Cedar trees.

Fake (adulterated) honey used in this study was collected fromApis mellifera. Hives were fed with sugar (sucrose) syrup thus sugarwas incorporated into the honey via bee-feeding. Study in this fieldhas shown that adulteration is also possible via bee-feeding syrupsand this can cause chemical modifications of the honey qualitysimilar to artificial adulteration via direct syrup incorporation tohoney (Cordella, Militão, Clément, Drajnudel, & Cabrol-Bass, 2005).

Maple syrup produced from the xylem sap of maple tree andcontains primarily sucrose and water. Maple syrups retrieved tothe study are Canadian origin and were used as a non-honey con-trol group. Fructose syrup was directly purchased from the market.Grape molasses (pekmez) is traditional syrup produced by boilingof the pressed grape juice and special grape soil mixture or creamof lime. It is rich in both carbohydrates and minerals.

2.2. Instrumentation and sample analysis

Spectra from all samples were collected in the one-bounce ATRmode in a Spectrum 100 FTIR spectrometer (Perkin-Elmer Inc.,Norwalk, CT, USA) equipped with a Universal ATR accessory.Samples were placed on Diamond/ZnSe crystal plate (Perkin-Elmer) and scanned from 4000 to 650 cm�1 for 50 scans withresolution of 4 cm�1 at room temperature. Each sample was repli-cated three times. Identical spectra were obtained in each case. Thisprocess was done to see the accuracy of the absorbance values,which might be affected from intra-sample variability and fromvariation in experimental conditions. Average spectra were usedfor further analysis. Data manipulations were carried out viaSpectrum 100 software (Perkin-Elmer).

2.3. Chemometrics

Cluster and Principal Component Analysis were applied to clas-sify the samples based on spectral differences. For the determina-tion of spectral differentiation among studied groups, clusteranalysis was performed via OPUS 5.5 software (Bruker Optics,GmbH). Vector normalised, first derivative of each spectrum inthe range of 1800–750 cm�1 was used as an input data. Spectraldistances were calculated between pairs of spectra as Pearson’scorrelation coefficients and Euclidean distance was used to calcu-late the sample similarities and to indicate the complete linkageclustering by Ward’s algorithm.

Principle Component Analyses (PCA) was used as a data reduc-tion method where each spectrum, which consists of hundreds of

Page 3: Differentiation of Anatolian honey samples from different botanical ...

236 S. Gok et al. / Food Chemistry 170 (2015) 234–240

absorbance values, is represented by a point in a multidimensionalspace using a linear transformation.

In this work, PCA was conducted on the ATR-FTIR spectraover 4000–650 cm�1 1700–1600 cm�1, 1175–940 cm�1 and 940–700 cm�1 range using by ‘‘Kinetics’’, a custom made programrunning under MATLAB (Matlab, Mathworks Inc.).

3. Results & discussion

In the current study, ATR-FTIR spectroscopy has been used tocompare honey samples based on their spectral differences in the4000–650 cm�1 spectral region. A representative ATR-FTIR spec-trum of honey is given in Fig. 1. Table 1 presents the band assign-ments along with the corresponding modes of vibrations in theATR-FTIR spectrum of honey, based on the literature (Gallardo-Velázquez, Osorio-Revilla, Loa, & Rivera-Espinoza, 2009; Kelly,Downey, & Fouratier, 2004; Movasaghi, Rehman, & Rehman,2008; Sivakesava & Irudayaraj, 2001; Subari et al., 2012; Tewari& Irudayaraj, 2004, 2005).

Fig. 2 shows comparative infrared spectra of all samples in the4000–650 cm�1 region. In this figure, spectral differences betweenthe groups were clearly seen.

Based on the spectral differences Hierarchical Cluster Analysis(HCA) and PCA have been applied to different spectral regions.Use of chemometrics together with classical methods for the

Fig. 1. Representative ATR-FTIR spectrum of ho

Table 1General band assignment of ATR-FTIR spectrum of honey. The related references are indic

Region 1 3000–2800 cm�1 C–H stretching (carbohydrates) (GallardoO–H stretching (carboxylic acids) (MovasNH3 stretching (free amino acids) (Gallard

Region 2 1700–1600 cm�1 O–H stretching/bending (water) (Cai & SiC@O stretching (mainly from carbohydraN–H bending of amide I (mainly proteins

Region 3 1540–1175 cm�1 O–H stretching/bending (Gallardo-VelázqC–O stretching (carbohydrates) (Tewari &C–H stretching (carbohydrates) (Tewari &C@O stretching of ketones (Tewari & Irud

Region 4 1175–940 cm�1 C–O & C–C stretching (carbohydrates) (SuRing vibrations (mainly from carbohydrat

Region 5 940–700 cm�1 Anomeric region of carbohydrates (MathlC–H bending (mainly from carbohydratesRing vibrations (mainly from carbohydrat

classification of different honey samples has been proposed in pre-vious researches. In 1960, discriminant functions of monosaccha-ride and ash content in addition to pH values were used forclassification of honey samples (Kirkwood, Mitchell, & Smith,1960). Linear discriminant analysis was employed to select mostuseful measurands by evaluating different sugars, water, pH value,colour, diastase enzyme activity conductivity and hyroxymethyl-furfural content. Later, by using pH value, free acidity, electricalconductivity, fructose, glucose and raffinose contents, botanicalorigins of honeys were estimated perfectly (Devillers, Morlot,Pham-Delegue, & Dore, 2004). In addition, flower honey was char-acterised by high concentration values for glucose and fructose andlow free acidity, polyphenol content, lactone quantity and electri-cal conductivity; whereas honeydew honeys have low concentra-tion glucose and fructose while showing high free acidity,polyphenol content, lactone quantity and electrical conductivity(Sanz, Gonzalez, de Lorenzo, Sanz, & Martinez-Castro, 2005).

The algorithms behind the cluster and Principal ComponentAnalysis that were used in the current study are quite different.PCA-like techniques can be preferred primarily for the determina-tion of general relationship among data (Gasper et al., 2010). How-ever, if one wants to show the grouping of similar data gatheredfrom different samples, cluster analysis must be performed(Wang & Mizaikoff, 2008). Similar samples tend to be classifiedin the same cluster and the level of difference between the clusters

ney in the 4000–650 cm�1 spectral region.

ated in the parenthesis.

-Velázquez et al., 2009)aghi et al., 2008)o-Velázquez et al., 2009; Sivakesava & Irudayaraj, 2001)

ngh, 2004; Stuart, 1997)tes) (Gallardo-Velázquez et al., 2009)) (Philip, 2009)uez et al., 2009; Tewari & Irudayaraj, 2004)Irudayaraj, 2004)Irudayaraj, 2005)

ayaraj, 2004)bari et al., 2012; Tewari & Irudayaraj, 2005)es) (Gallardo-Velázquez et al., 2009; Tewari & Irudayaraj, 2004)outhi & Koenig, 1986; Subari et al., 2012)) (Gallardo-Velázquez et al., 2009; Kelly et al., 2004; Tewari & Irudayaraj, 2004)es) (Tewari & Irudayaraj, 2004)

Page 4: Differentiation of Anatolian honey samples from different botanical ...

Fig. 2. Comparative ATR-FTR spectra of all samples in the 4000–650 cm�1 spectral region. Spectra were normalised to the band located at 3300 cm�1.

Fig. 3. Hierarchical clustering of all samples in the 1800–750 cm�1 (fingerprint) spectral region.

S. Gok et al. / Food Chemistry 170 (2015) 234–240 237

is indicated with heterogeneity values (Ward, 1963). Ward’s algo-rithm was previously reported to give one of the best predictions,among the different methods used in cluster analysis (Lasch,Haensch, Naumann, & Diem, 2004; Severcan, Bozkurt, Gurbanov,& Gorgulu, 2010). As opposed to other methods, algorithm triesto find groups which are as homogeneous as possible. This means

that only two groups, which show the smallest growth in heteroge-neity factor H, are merged. Detailed information about this methodwas reported in Severcan et al. (2010).

In this study, in order to reduce the number of variables prior toperforming cluster analysis, we used the PCA. This was conductedon four different regions. Depending on the PCA outputs, the

Page 5: Differentiation of Anatolian honey samples from different botanical ...

Fig. 4. PCA scatter plots for all of the samples over 4000–650 cm�1 (A), 1700–1600 cm�1 (B), 1175–940 cm�1 (C), and 940–700 cm�1 (D) spectral region. (Ellipses have aconfidence factor of 0.8.)

238 S. Gok et al. / Food Chemistry 170 (2015) 234–240

regions having the highest principal component (PC) values wereselected for HCA. Other spectral regions, used for PCA analysis,were also tried for hierarchical clustering of the sample groups.However, the best differentiation was achieved only in the1800–750 cm�1 region. This region includes the anomeric regionat 950–750 cm�1 which was frequently preferred for the spectralanalysis of carbohydrates in IR spectroscopy. Analysis in this rangemakes it possible to distinguish bands characteristic for a and bconformers or pyranoid and furanoid ring vibrations of monoand polysaccharides (Mathlouthi & Koenig, 1986). In addition toalpha and beta conformers, the fingerprint region (1800–750 cm�1) contains other contributions that arise from differentmolecules. Especially water (around 1640 cm�1) and minuteamount of protein molecules give bands in the indicated region.Also the differences among honey samples can be related not onlyto different water content in the different honeys but also to theinteraction between water molecules and carbohydrates, depend-ing on their structure. The precise assignment of bands in thisregion cannot be stated unequivocally. However fingerprint regionprovides a unique spectrum for each compound where theposition and intensity of the bands are specific for everypolysaccharide (Filippov, 1992; Li-Chan et al., 2010). Therefore,

1800–750 cm�1 spectral region was selected for successfuldiscrimination of clusters.

For the calculation of sample similarities, the Euclidean distancewas used indicating the complete linkage clustering values. Theresults obtained are represented in Fig. 3 in the form of dendro-grams. Clear cut classes were gathered over the range of 1800–750 cm�1 with high heterogeneity values (up to 10). All of the treeoriginated samples (chestnut, cedar, pine) are aggregated in onecluster on the left arm. As the maple syrup is also the maple treeoriginated sample, it shows more similarity to tree originatedgroup than to the flower originated ones. One arm of the secondcluster is composed of flower originated honey samples includingpolyfloral, anzer, organic, Taurus flower and rhododendron honeys.Anzer, organic and Taurus flower honeys are region specific sam-ples, it is known that the purity of their botanical origins is higherthan polyfloral honey group. So they were clustered in the samearm. As the rhododendron honey is collected from the Black SeaRegion mountains, it was clustered closer to that group thanpolyfloral ones. Fructose syrup, grape molasses and the fake(adulterated) honey were agglomerated on the far right arm ofthe second cluster in that they differ from the natural samples interms of their carbohydrate content significantly.

Page 6: Differentiation of Anatolian honey samples from different botanical ...

Fig. 5. The first 4 eigenvectors (PCA loading spectra) for the spectral region 1175–940 cm�1.

S. Gok et al. / Food Chemistry 170 (2015) 234–240 239

Based on spectral differences a mean-centered PCA wasconducted to all samples on the infrared spectra over the range of4000–650 cm�1 (Fig. 4A), 1700–1600 cm�1 (Fig. 4B), 1175–940 cm�1 (Fig. 4C) and 940–700 cm�1 (Fig. 4D). Distinct segregationand clustering between the samples were apparent in all figures.Samples were grouped close together creating uniform clustersfor each of the analysed honey types in the PCA scatter plot.

PCA is a data reduction method where each spectrum, whichconsists of hundreds of absorbance values, is represented by apoint in a multidimensional space using a linear transformation.The coordinates of the point are the principal components (PC)and the plot obtained is called the scores plot. The transformationmatrix in PCA consists of the eigenvectors of the covariance matrixof the whole set of spectra. Thus, each spectrum can be representedas the sum of a number of weighted orthonormal eigenvectors,where the weights are the PCs. In PCA the eigenvectors are orderedsuch that corresponding eigenvalues appear in decreasing order.Since the eigenvalues represent the variance of the correspondingeigenvectors, the first eigenvector describes the highest part of var-iability; the second eigenvector describes the second largest sourceof variability, and so on. Therefore, the corresponding scores defin-ing a specific spectrum indicate the amount of contribution of eacheigenvector to this spectrum. An important property of PCA is thatno other orthogonal transformation can give smaller error (inmean squared sense) than that of PCA if the number of transformcoefficients is truncated at some value. In practice, the first 3 or4 PCs are sufficient to represent an FTIR spectrum and observethe significant differences among spectra. Therefore, scores of sim-ilar spectra in a multidimensional score plot are clustered. On theother hand, as the dissimilarity of clusters increases the clustersare separated from each other.

PCA was applied to FTIR spectra of all groups, obtaining an evi-dent discrimination (score plot) of the different pure honey sam-ples with respect to the fake honey, grape molasses, fructose andmaple syrups (Fig. 4A). In practice, it is more convenient to plot2-D plots of any two of the PCs. Such a plot is the projection of amultidimensional plot onto a 2-D space. Therefore it may be possi-ble to have overlaps of clusters in one plot and one has to checkother combinations of PCs in other plots to observe cluster separa-tions. Here, successful significant differentiation of all investigatedgroups has been achieved in a single 2D plot. Only rhododendronsamples make penetrations to both tree and flower originatedgroups. This result is in consistence with the fact that rhododen-dron can be classified as brier, therefore, located closer to both treeand flower originated samples. Infrared modes of water are veryintense and may overlap with the carbohydrate modes. The majorinfrared bands of water located at 3920 cm�1, 3490 cm�1 and3280 cm�1 for O–H stretching and 1645 cm�1 for H–O–H bendingvibrations (Stuart, 1997). In the scope of this study, primary goalis the discrimination of honey samples coming from different floralorigins rather than giving numerical information about the constit-uents of samples. Also as it can be seen in the general honey spec-trum (Fig. 1), vibrations coming from bulk water at 2250 cm�1 arenegligible. So the water in the samples is mainly bounded water.Water content itself can be used as a parameter for honey charac-terisation (Manikis & Thrasyvoulou, 2001; Persano Oddo & Piro,2004). Thus, vibrations coming from water may also contributethe discrimination success in 4000–650 cm�1 region, in additionthe differences in carbohydrate content and structure of samples.

A clear splitting of the data can be observed as depicted inFig. 4B, by the first two principal components in the 1700–1600 cm�1 region. This describes, 97.1% of the total variance forPC1 and 2.4% for PC2. Studies revealed that pollen proteins canbe used as a marker for botanical classification of honey. At leastnineteen different protein bands were visualised by SDS–PAGEexperiments in honeys of different botanical origin; molecular

weights of proteins in honey can vary depending on the honeybeespecies (Won, Lee, Ko, Kim, & Rhee, 2008). Immunological charac-terisation of honey major protein and its application have beenreported by Won, Li, Kim, and Rhee (2009). The bands in the region1700–1600 cm�1 had been previously assigned as amide I vibra-tions of the honey proteins (Philip, 2009). However, water mole-cules give strong absorption between 1640 and 1650 cm�1 (Cai &Singh, 2004) so the discrimination in this region can be explainedby the difference in water content, protein content and water–carbohydrate interactions between sample groups.

PCA results of the regions lying between the 1175–940 cm�1

and 940–700 cm�1 are given in Fig. 4C and D, respectively. Herehighly successful discrimination of all samples is achieved due todifferences in their carbohydrate content and structure. Inaddition, tree originated samples were separated from floweroriginated samples clearly, in both figures.

The peaks observed in the loading spectra explain the chemicalbasis of the discrimination between different types of honey(Fig. 5). As mentioned above, PCA provides a decomposition of anFTIR spectrum X in terms of a set of eigenvectors (or PCA loadingvectors/spectra) Vi, as in X =

PixiVi, where xi are the score values.

This implies that for positive score values of xi positive peaks,and for negative score values of xi negative peaks of the PCA load-ing spectra Vi have significant contribution to the spectrum X. Forexample, using the PCA results for the interval 1175–940 cm�1

and the corresponding eigenvectors shown in Fig. 5, it can be seenthat chestnut honey is characterised by a high score value on PC1(which explains 79.7% of the total variance). PC1 has a strong con-tribution from the line at around 1020 cm�1 (carbohydrate band).This contribution is particularly high in chestnut honey. However,using only two PCs in general cannot provide a complete represen-tation and contributions of other loading spectra may need to betaken into account to see how significant the observed contribu-tion of the peak is.

4. Conclusion

The results from the current study point out that there aremany considerable variations in the spectral parameters of honeysamples which come from different botanical origins. Especially

Page 7: Differentiation of Anatolian honey samples from different botanical ...

240 S. Gok et al. / Food Chemistry 170 (2015) 234–240

the variations in the water, carbohydrates and protein dominantlydetermine the specific spectral pattern of each sample. ATR-FTIRspectroscopy in combination with multivariate analysis enablesthe extraction of useful quantitative information via single rapidmeasurement for classification of botanical origin of honey sam-ples. Based on spectral variations, successful differentiation wasobtained with HCA and PCA. Results of this study exposed thepotential power of ATR-FTIR spectroscopy in automated and highlysensitive botanical origin estimation of honey. It should be empha-sized that some of the natural honeys (i.e., chestnut and Anzer) arerelatively expensive. For instance in Turkey, the price of pine,flower, chestnut, rhododendron and Anzer honeys increase in thegiven order. As an example, the cost of pine honey is 20–30 $/kgand Anzer is 700 $/kg. In developing counties, due to the high costof such honeys, adulterated honeys or fake honeys are put on themarkets or even open markets with or without a brand. It is there-fore very important to discriminate between natural and adulter-ated and/or fake honey types in developing countries, as in Turkey.

Acknowledgement

We would like to thank Mrs. Enise Gulsum Su, for supplying theTaurus flower honey samples (from Fethiye region) used in thisstudy.

References

Bryant, V. M., Jr., & Jones, G. D. (2001). The R-values of honey: Pollen coefficients.Palynology, 25, 11–28.

Cai, S., & Singh, R. B. (2004). A distinct utility of the amide III infrared band forsecondary structure estimation of aqueous protein solutions using partial leastsquares methods. Biochemistry, 43, 2541–2549.

Chalmers, J. M., & Griffiths, P. R. (2002). Handbook of vibrational spectroscopy (Vol. 5).Chichester: John Wiley & Sons.

Chung, H., Ku, M. S., & Lee, J. S. (1999). Comparison of near-infrared and mid-infrared spectroscopy for the determination of distillation property of kerosene.Vibrational Spectroscopy, 20(2), 155–163.

Cordella, C., Militão, J. S. L. T., Clément, M. C., Drajnudel, P., & Cabrol-Bass, D. (2005).Detection and quantification of honey adulteration via direct incorporation ofsugar syrups or bee-feeding: Preliminary study using high-performance anionexchange chromatography with pulsed amperometric detection (HPAEC-PAD)and chemometrics. Analytica Chimica Acta, 531(2), 239–248.

Cordella, C., Moussa, I., Martel, A. C., Sbirrazzuoli, N., & Lizzani-Cuvelier, L. (2002).Recent developments in food characterization and adulteration detection:Technique-oriented perspective. Journal of Agricultural Food Chemistry, 50,1751–1764.

Devillers, J., Morlot, M., Pham-Delegue, M. H., & Dore, J. C. (2004). Classification ofmonofloral honeys based on their quality control data. Food Chemistry, 86,305–312.

Etzold, E., & Lichtenberg-Kraag, B. (2008). Determination of the botanical origin ofhoney by Fourier-transformed infrared spectroscopy: An approach for routineanalysis. European Food Research and Technology, 227, 579–586.

Filippov, M. P. (1992). Practical infrared spectroscopy of pectic substances. FoodHydrocolloids, 6(1), 115–142.

Gallardo-Velázquez, T., Osorio-Revilla, G., Loa, M. Z., & Rivera-Espinoza, Y. (2009).Application of FTIR-HATR spectroscopy and multivariate analysis to thequantification of adulterants in Mexican honeys. Food Research International,42, 313–3318.

Gasper, R., Mijatovic, T., Bénard, A., Derenne, A., Kiss, R., & Goormaghtigh, E. (2010).FTIR spectral signature of the effect of cardiotonic steroids with antitumoralproperties on a prostate cancer cell line. Biochimica et Biophysica Acta (BBA) –Molecular Basis of Disease, 1802(11), 1087–1094.

Kayacier, A., & Karaman, S. (2008). Rheological and some physicochemicalcharacteristics of selected Turkish honeys. Journal of Texture Studies, 39, 17–27.

Kelly, J. F. D., Downey, G., & Fouratier, V. (2004). Initial study of honey adulterationby sugar solutions using mid-infrared (MIR) spectroscopy and chemometrics.Journal of Agricultural Food Chemistry, 52, 33–39.

Kirkwood, K. C., Mitchell, T. J., & Smith, D. (1960). An examination of the occurrenceof honeydew in honey. Analyst, 85, 412–416.

Kucuk, M., Kolayli, S., Karaoglu, S., Ulusoy, E., Baltaci, C., & Candan, F. (2007).Biological activities and chemical composition of three honeys of different typesfrom Anatolia. Food Chemistry, 100(2), 526–534.

Lasch, P., Haensch, W., Naumann, D., & Diem, M. (2004). Imaging of colorectaladenocarcinoma using FTIR microspectroscopy and cluster analysis. BiochimicaBiophysica Acta, 1688, 176–178.

Li-Chan, E., Chalmers, J. M., & Griffiths, P. (2010). Applications of vibrationalspectroscopy in food science. 1 Instrumentation and fundamental applications.Chichester: John Wiley & Sons.

Lichtenberg-Kraag, B., Hedtke, C., & Bienefeld, K. (2002). Infrared spectroscopy inroutine quality analysis of honey. Apidologie, 33(3), 327–337.

Lindner, P., Bermann, E., & Gamarnik, B. (1996). Characterization of citrus honey bydeuterium NMR. Journal of Agricultural Food Chemistry, 44, 139–140.

Low, N. H., & South, W. (1995). Determination of honey authenticity by capillary gaschromatography. Journal of AOAC International, 78, 1210–1218.

Manikis, I., & Thrasyvoulou, A. (2001). The relation of physico-chemicalcharacteristics of honey and the crystallization sensitive parameters. Apiacta,36, 106–112.

Mathlouthi, M., & Koenig, J. L. (1986). Vibrational spectra of carbohydrates. Advancesin Carbohydrate Chemistry and Biochemistry, 44, 7–89.

Movasaghi, Z., Rehman, S., & Rehman, I. (2008). Fourier transform infrared (FTIR)spectroscopy of biological tissues. Applied Spectroscopy Reviews, 43(2), 134–179.

Onat, F. Y., Yegen, B. C., Lawrence, R., Oktay, A., & Oktay, S. (1991). Mad honeypoisoning in man and rat. Reviews on Environmental Health, 1, 4–9.

Persano Oddo, L., & Piro, R. (2004). Main European unifloral honeys: Descriptivesheets. Apidologie, 35, 38–81 (special issue).

Philip, D. (2009). Honey mediated green synthesis of gold nanoparticles.Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 73(4),650–653.

Radovic, B. S., Goodacre, R., & Anklam, E. (2001). Contribution of pyrolysis massspectrometry. (Py-MS) to authenticity testing of honey. Journal of Analytical andApplied Pyrolysis, 60, 79–87.

Rios-Corripio, M. A., Rios-Leal, E., Rojas-López, M., & Delgado-Macuil, R. (2011). FTIRcharacterization of Mexican honey and its adulteration with sugar syrups byusing chemometric methods. Journal of Physics: Conference Series, 274, 012098.

Rios-Corripio, M. A., Rojas-López, M., & Delgado-Macuil, R. (2012). Analysis ofadulteration in honey with standard sugar solutions and syrups usingattenuated total reflectance-Fourier transform infrared spectroscopy andmultivariate methods. CYTA – Journal of Food, 10(2), 119–122.

Ruoff, K. (2006). Authentication of the botanical origin of honey (Doctoraldissertation). Swiss Federal Institute of Technology, Zurich.

Sanz, M. L., Gonzalez, M., de Lorenzo, C., Sanz, J., & Martinez-Castro, I. A. (2005).Contribution to the differentiation between nectar honey and honeydew honey.Food Chemistry, 91, 313–317.

Senyuva, H. Z., Gilbert, J., Silici, S., Charlton, A., Dal, C., Gurel, N., et al. (2009).Profiling Turkish honeys to determine authenticity using physical and chemicalcharacteristics. Journal of Agricultural and Food Chemistry, 57(9), 3911–3919.

Severcan, F., Bozkurt, O., Gurbanov, R., & Gorgulu, G. (2010). FT-IR spectroscopy indiagnosis of diabetes in rat animal model. Journal of Biophotonics, 3(8–9),621–631.

Silici, S., & Gokceoglu, M. (2007). Pollen analysis of honeys from Mediterraneanregions of Anatolia. Grana, 46, 57–64.

Sivakesava, S., & Irudayaraj, J. (2001). Prediction of inverted cane sugar adulterationof honey by Fourier transform infrared spectroscopy. Journal of Food Science, 66,972–978.

Stuart, B. (1997). Biological applications of infrared spectroscopy. Chichester, UK:ACOL Series, Wiley.

Subari, N., Saleh, J. M., Shakaff, A. Y. M., & Zakaria, A. (2012). A hybrid sensingapproach for pure and adulterated honey classification. Sensors, 12,14022–14040.

Swallow, K. W., & Low, N. H. (1990). Analysis and quantitation of the carbohydratesin honey using high-performance liquid chromatography. Journal of AgriculturalFood Chemistry, 38, 1828–1832.

Tewari, J., & Irudayaraj, J. (2004). Quantification of saccharides in multiple floralhoneys using Fourier transform infrared microattenuated total reflectancespectroscopy. Journal of Agricultural Food Chemistry, 52, 3237–3243.

Tewari, J. C., & Irudayaraj, J. M. K. (2005). Floral classification of honey using mid-infrared spectroscopy and surface acoustic wave based z-Nose sensor. Journal ofAgricultural Food Chemistry, 53, 6955–6966.

TUIK (Turkish Statistical Institute). (2013). _Istatistiksel Göstergeler.Wang, L., & Mizaikoff, B. (2008). Application of multivariate data-analysis

techniques to biomedical diagnostics based on mid-infrared spectroscopy.Analytical and Bioanalytical Chemistry, 391(5), 1641–1654.

Ward, J. H. (1963). Hierarchical grouping to optimize an objective function. Journalof American Statistical Association, 58, 236.

White, J. W., Winters, K., Martin, P., & Rossmann, A. (1998). Stable carbon isotoperatio analysis of honey: Validation of internal standard procedure forworldwide application. Journal of AOAC International, 81, 610–619.

Won, S.-R., Lee, D.-C., Ko, S. H., Kim, J.-W., & Rhee, H.-I. (2008). Honey major proteincharacterization and its application to adulteration detection. Food ResearchInternational, 41, 952–956.

Won, S.-R., Li, C.-Y., Kim, J.-W., & Rhee, H.-I. (2009). Immunological characterizationof honey major protein and its application. Food Chemistry, 113(4), 1334–1338.


Recommended