Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts

Transcription

1 Cent. Eur. J. Chem. 12(6) DOI: /s Central European Journal of Chemistry Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts Research Article Joanna Ronowicz *, Bogumiła Kupcewicz, Elżbieta Budzisz Department of Inorganic and Analytical Chemistry, Faculty of Pharmacy, Collegium Medicum, Nicolaus Copernicus University, Bydgoszcz, Poland Received 16 August 2013; Accepted 8 January 2014 Abstract: In this study, antioxidant properties of commercial green teas and dietary supplements containing Camellia sinensis extracts were evaluated. Extracts were examined using two antioxidant assays (DPPH radical method and ABTS + cation radical method). A Folin Ciocalteu assay was used to evaluate the total polyphenol content in the extracts. In order to compare and characterize the investigated Camellia sinensis extracts, chemometric techniques based on fingerprint chromatograms, antioxidant activity and total polyphenol content were applied. Application of chemometric methods allowed for reduction of multidimensionality of the data set and grouped the samples into differentiable clusters. The relationship between the antioxidant activity and total polyphenol content was also assessed. The results indicated that extracts with the higher polyphenolic content exhibited the stronger antiradical activity against both DPPH radicals and ABTS + cation radicals. The multivariate calibration technique (such as a tree regression algorithm) can be a useful tool for rapid determining the antioxidant activity of a herbal product based on its fingerprint chromatogram. Keywords: Antioxidant activity Camellia sinensis extract Chemometric methods Total polyphenol content Versita Sp. z o.o. 1. Introduction Multi-composition of bioactive compounds in green tea is responsible for the increasing interest in this plant material. Green tea, as a rich source of natural antioxidants, protects cells against damages caused by reactive oxygen species and free radicals. Significant health properties of Camellia sinensis extract, which are considered to be closely related to its antioxidant activity, have been summarized in several research papers [1-4]. Green tea reduces the risk of cardiovascular diseases, certain types of cancer, cataract formation, neurodegenerative and atherosclerotic disorders. Green tea extract has also a protective effect against inflammation caused by lipid peroxidation and excessive free radical production. These beneficial effects on human health result from the antioxidant properties of the polyphenolic compounds present in Camellia sinensis extract. This extract indicates a high content of active flavan-3-ols such as epigallocatechin-3-gallate, epigallocatechin, epicatechin-3-gallate and epicatechin [5-6]. The content of active compounds depends on harvest time, method of drying, grinding, storage conditions and it determines the quality of green tea. The present study is focused on the determination and comparison of total polyphenol content and antioxidant properties of commercial green teas and dietary supplements containing an extract from the leaves of Camellia sinensis. In this work, attention is directed toward chemometric methods as potentially useful tools for evaluation of antioxidant activity of herbal products. The aim of the study was also to construct a regression model providing the antioxidant activity of the extract based on the fingerprint chromatogram. Antioxidant activity was treated as a measure of the quality of the extract which directly affects its therapeutic efficacy. The objective of the project was not to identify and quantify each ingredient present in the studied * joanna.ronowicz@cm.umk.pl 700

2 J. Ronowicz, B. Kupcewicz, E. Budzisz extracts. A chromatographic fingerprinting technique was applied. This approach provides a chromatographic pattern of the extract [7] which presents multiple chemical compounds and their relative concentrations. In the literature, various chemometric methods for evaluation of the quality and authentication of herbal medicines have been reported [8]. These chemometric techniques are combined with different analytical methods, including chromatographic (HPLC, GC and CE) and spectroscopic (NMR, IR, UV) analysis. Xie et al. [9] described the application of unsupervised pattern recognition techniques in similarity analysis based on fingerprints for evaluation of the quality consistency of herbal medicines. The chromatographic fingerprint itself provides no information about the bioactivity and thus the therapeutic efficacy of herbal products. That is why the bioactivity of herbal extract should be correlated with its chromatographic profile by means of multivariate calibration techniques [10]. The prediction of green tea antioxidant capacity from fingerprint chromatograms was described by several research groups [11-14]. The linear multivariate calibration techniques were used to predict the total antioxidant capacity of green tea from chromatographic fingerprints [11]. The orthogonal projections to latent structures (O-PLS) as a modified version of ordinary partial least squares (PLS) was applied for prediction the antioxidant capacity of green tea from dissimilar chromatographic fingerprints [12]. Nederkassel et al. [13] used partial least squares (PLS) and uninformative variable elimination partial least squares (UVE-PLS) to construct a multivariate regression model in order to predict the total antioxidant capacity from the fast chromatograms. Daszykowski et al. [14] constructed the robust partial least squares model to predict the total antioxidant capacity of green tea extracts. This model had the same fit and predictive power as the classical partial least squares model. Chen et al. [15] used a support vector machine (SVM)- based model and a principle component regression (PCR)-based model for estimating the pharmacological effect of herbal medicines from the chromatographic peak areas. Tistaert et al. [16-18] modeled the antioxidant activity of the samples as a function of the fingerprints using the orthogonal projections to latent structures (O-PLS) technique. The regression coefficients of the models were studied to indicate the peaks potentially responsible for the antioxidant activity. Apart from the various algorithms mentioned above, the artificial neural network (ANN) algorithm was also used for prediction of antioxidant capacity [19]. The multilayer perceptron as the most popular network architecture was applied. The obtained model indicated good predictive abilities but the relationship between the predictor variables and the predicted output value (antioxidant capacity) was not easy to interpret and for this reason, the ANN model is considered a black box. In this study, in contrast to the above-mentioned papers, a tree regression algorithm was used to build a predictive model. The advantages of the tree regression model are simplicity and good interpretability of the compounds contribution to the antioxidant capacity. Due to the clear set of decision rules, it is possible to indicate the chromatographic peaks which are responsible for the antioxidant activity of the extract. Thus, besides predictive abilities, the tree regression model has also good describing abilities. This chemometric approach can be considered as a fast way for determining the total antioxidant capacity of herbal products based on their fingerprint chromatograms. This antioxidant activity should be considered as an important indicator in evaluating the quality of herbal products and thus their effectiveness in the prevention of diseases caused by free radicals and reactive oxygen species. 2. Experimental procedure 2.1. Materials In this work, six commercial green teas (marked with symbols GT1 to GT6) and six dietary supplements (marked with symbols DS1 to DS6) containing the extract from the leaves of Camellia sinensis were analyzed. A description of investigated herbal products is given in Table HPLC/DAD analysis The green tea extracts were analyzed by reversed phase high performance liquid chromatography (RP- HPLC) to obtain fingerprint chromatograms which are specific chemical patterns of plant materials. The optimal separation of the polyphenolic compounds present in the extracts was carried out using a GraceSmart C-18 column ( mm, 5 μm). The gradient elution was applied. The mobile phase consisted of 0.1% formic acid (A) and acetonitrile (B). The flow rate was 0.5 ml min -1 and the injection volume was 20 ml. The gradient program was as follows: 0-4 min, isocratic 17% B; 4-10 min, linear gradient 17-23% B; min, isocratic 23% B; min, linear gradient 23-17% B. The chromatograms were recorded at the wavelength 270 nm. The fingerprint chromatograms of the investigated Camellia sinensis extracts are depicted in Fig

3 Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts Table 1. The investigated commercial green teas and dietary supplements containing the extract from green tea leaves. Code Composition of the product indicated on the label by the manufacturer Product category: dietary supplement dosage form: DS1 fragmented green tea leaves capsules DS2 extract from green tea leaves (highly concentrated) capsules DS3 green tea extract tablets DS4 extract from green tea leaves (highly concentrated) capsules DS5 standardized green tea extract capsules DS6 dry extract from green tea leaves capsules GT1 GT2 GT3 GT4 GT5 GT6 Product category: green tea food product leaf green tea leaf green tea leaf green tea pressed green tea leaf green tea leaf green tea Sample preparation Each green tea sample was crushed in a mortar to make homogeneous powder. In the case of dietary supplements, the content of 10 tablets or 10 capsules were powdered in a porcelain mortar. One hundred miligrams of each powdered sample was accurately weighed into a glass vial, then 10 ml of methanol : water (80:20, v/v) was added and it was vortexed for 2 minutes. After that, the sample vial was sonicated in an ultrasonic bath at a temperature of 25 C for 15 minutes and then filtered through 0.22 μm membrane. One milliliter of the obtained solution was taken and mixed with 2 ml of methanol : water (50:50, v/v) in a glass vial. Then, 20 μl of sample solution was injected into the HPLC column. Three replicates of each extract were performed. Methanol and acetonitrile were of HPLC gradient grade from POCH (Gliwice, Poland). Ultrapure water from Milli-Q system (Millipore, Bedford, USA) was used in all experiments. The other reagents were of analytical grade Determination of total phenolic content Total phenolic content (TPC) in the studied extracts was determined by means of spectrophotometric method Figure 1. The fingerprint chromatograms of the investigated commercial green teas (marked with symbols GT1 to GT6) and dietary supplements containing Camellia sinensis extract (marked with symbols DS1 to DS6), registered at the wavelength of 270 nm. 702

4 J. Ronowicz, B. Kupcewicz, E. Budzisz with the use of Folin-Ciocalteu reagent, according to the method described by Djeridane et al. and Dóka et al. [20,21].This method is based on a polyphenols oxidation reaction by two strong inorganic oxidants (phosphomolibdic acid and phosphotungstic acid). A calibration curve was prepared for gallic acid as a standard. Total polyphenols, determined in methanolic extracts, were converted to gallic acid based on the equation of the calibration curve (r 2 = ; y = x ; where y is the absorbance and x is the gallic acid concentration in μg per 10 ml). The total polyphenol content of Camellia sinensis extracts was determined as equivalent of the gallic acid standard and was expressed as mg GAE/100 mg of dry material. The total polyphenol content in the studied extracts was determined in triplicate. The sample solution for this assay was prepared in the same manner as described in section Then 60 μl of a sample solution was transferred into a 10 ml volumetric flask and mixed with 0.5 ml of Folin-Ciocalteu reagent. After 5 min, 2 ml of sodium carbonate solution (20 g per 100 ml) was added to the mixture and then filled up with deionized water to 10 ml. After 30 minutes of incubation in the darkness at room temperature, the reaction mixture absorbance was measured at the wavelength of 750 nm. A detailed description of chemical principles of this assay can be found in reference [22] DPPH radical scavenging activity assay The radical-scavenging activity of Camellia sinensis extracts was determined on the basis of the scavenging activities of the stable 2,2-diphenyl-1-picrylhydrazyl (DPPH ) free radical as described by Molyneux and Pellati et al. [23,24] with slight modifications. The antiradical activity profile was plotted for each herbal product. This profile was used to determine the IC 50 value, which is defined as the concentration of the extract (μg ml -1 ) that causes the inactivation of 50% of the radicals present in a sample. Then 0.15 ml of each green tea extract at different concentrations was mixed with 1.0 ml DPPH methanolic solution (10-4 mol L -1 ). After 30 minutes of incubation in the darkness at room temperature, the absorbance was measured at the wavelength of 517 nm. All measurements were done in triplicate. The DPPH radical-scavenging activity (%) was calculated using following equation: Scavenging effect (% Inhibition) = [(A 0 A t )/A 0 ] 100% where A 0 is the absorbance of a blank sample (DPPH solution) and A t is the absorbance of a tested sample (DPPH solution after addition of Camellia sinensis extract). The antioxidant activity was expressed as the IC 50 value. This value was determined from the plotted graphs of scavenging activity against the concentration of the extract. A detailed description of chemical principles of DPPH radical scavenging activity assay can be found in reference [22] ABTS + scavenging activity assay ABTS + scavenging activity assay was carried out according to the method reported by Cai et al. [25]. The ABTS + solution was prepared by mixing 7 mm ABTS and 2.45 mm potassium persulfate and incubating in the darkness at room temperature for 12 hours. Then the 2,2 -azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) cation radicals (ABTS + ) solution was diluted to obtain an absorbance of about 0.70 at 734 nm ml of each green tea extract at different concentrations was mixed with 1.0 ml of the obtained ABTS + solution. The absorbance of the mixture was measured at 734 nm after 30 minutes of incubation in the darkness at room temperature. All tests were carried out in triplicate. The ABTS + scavenging effect was calculated as follows: Scavenging effect (% Inhibition) = [(A 0 A t )/A 0 ] 100% where A 0 is the absorbance of a blank sample (ABTS + solution) and A t is the absorbance of a tested sample (ABTS + solution after addition of Camellia sinensis extract). The antioxidant activity was expressed as the half-maximal inhibition concentration (IC 50 ) which was calculated from regression equation, where the abscissa represented the concentration of extracts (μg ml -1 ) and the ordinate the percent of scavenging capacity from the test. A detailed description of chemical principles of ABTS + scavenging activity assay can be found in reference [22] Chemometric methods The collected data were subjected to chemometric analysis. All calculations were performed using STATISTICA 10.0 software (StatSoft, Tulsa, Oklahoma, USA) and Matlab software (Mathworks Inc., Natick, MA, USA). A data matrix contained the spectrophotometric (total phenolic content, DPPH and ABTS + scavenging activity) and chromatographic data. In this study, three unsupervised pattern recognition methods such as cluster analysis, k-means clustering algorithm and principal component analysis were applied. These are multivariate techniques for analysis of relationships and detection of some similarities and differences within large data sets. This work is focused on the various chemometric techniques as potentially useful tools for evaluation of Camellia sinensis extracts in terms of their antioxidant properties. Furthermore, 703

5 Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts Figure 2. The dendrogram obtained by hierarchical cluster analysis using Ward s method and squared Euclidean distance metric. in order to build a calibration model for predicting the antioxidant activity based on the fingerprint chromatogram, a tree regression algorithm was applied. A detailed description of chemometric methods used in this work can be found in references [26-30]. 3. Results and discussion 3.1. Chemometric similarity analysis The data set for chemometric analysis was formed by measurements of ABTS + scavenging activity assay, DPPH scavenging activity assay, total polyphenols content assay and chromatographic data. Before the chemometric analysis, the raw data were standardized by subtracting a sample mean from each variable value and dividing it by the standard deviation. This transformation procedure allowed to unify the effects of all variables on the picture of the relationships between the studied extracts. Various segmentation algorithms (hierarchical cluster analysis, non-hierarchical clustering method and principal component analysis) were applied in order to construct the descriptive models describing the similarities and differences between the investigated extracts in terms of their antioxidant properties. In the cluster analysis, Ward s method was applied for aggregation of the samples, the squared Euclidean distance was used as the distance measure. Thus, segments with minimal internal differentiation were obtained. The dendrogram derived from a cluster analysis is shown in Fig. 2. The clusters were defined based on the values of variables describing the investigated extracts. As can be seen, the extracts were classified into two various sub-groups according to their spectrophotometric measurements (TPC, DPPH and ABTS + assay) and their chromatographic profiles. The unsupervised hierarchical cluster analysis identified two major segments. One of them consisted of four dietary supplements (DS2, DS4, DS5, DS6), whereas the second cluster was formed by green teas (GT1, GT2, GT3, GT4, GT5, GT6) and two dietary supplements (DS1, DS3). The supplement marked with the symbol DS1 includes fragmented green tea leaves, which may explain its belonging to the cluster, which is dominated by leaf green teas. The supplement marked with symbol DS3, unlike the others dietary supplements, has a different dosage form (coated tablets). Non-hierarchical clustering algorithm (k-means method) with the use of v-fold cross-validation procedure allowed for iterative determination the most optimal number of clusters. Three clusters were clearly identified. The profile of each segment, including the results of analysis of antioxidant activity and polyphenol content, was plotted (Fig. 3). The extracts classified into the same cluster indicate a similar content of polyphenols and antioxidant properties. Like in the hierarchical cluster analysis, in k-means method four dietary supplements (DS2, DS4, DS5, DS6) formed one cluster which is characterized by a high antioxidant activity (a low IC 50 value for both DPPH radical and ABTS + radical) and a high content of polyphenols. Principal component analysis (PCA) was also applied. PCA is one of the most commonly used exploratory 704

6 J. Ronowicz, B. Kupcewicz, E. Budzisz Figure 3. Three clusters obtained by k-means clustering algorithm. Table 2. Factor loadings for spectrophotometric data. Factor 1 Factor 2 Factor 3 IC 50 ABTS IC 50 DPPH TPC method to reduce large complex data sets into a series of orthogonally related components of interpretable size. Each principal component is a linear combination of the original variables. Reducing the dimensionality of the original data by means of principal component analysis enables graphical illustration of the relationship between the extracts in terms of their antioxidant properties. The calculations gave three principal components as the most significant. These three factors explained the major part of the variability in the data set (over 77%). The first principal component explained most of the variance in the original data (over 50%). Analysis of factor loadings (Table 2) indicates that the first factor is highly associated with the ABTS + radical scavenging activity (expressed as IC 50 ABTS value), the DPPH radical scavenging activity (expressed as IC 50 DPPH value) and total polyphenols content (TPC), with high loadings of 0.916, and , respectively. Thus, the first principal component is strongly correlated with these variables and describes the antioxidant properties of the investigated extracts very well. Furthermore, it can be seen a negative correlation between the values of IC 50 and total polyphenols content. The extracts with the higher content of polyphenols indicate the lower values of IC 50, which means the stronger antiradical activity against both DPPH radicals and ABTS + cation radicals. 3D scores plot (Fig. 4) clearly shows distinct clustering of the investigated Camellia sinensis extracts. The principal component analysis, as well as the cluster analysis, identified two major clusters of the extracts, clearly differentiated in terms of antioxidant properties. Four dietary supplements (DS2, DS4, DS5, DS6) are characterized by a high antiradical activity and a high content of polyphenolic compounds. The other extracts belong to a second cluster, which is characterized by the lower antioxidant activity and the lower content of polyphenols. Application of various unsupervised pattern recognition methods gave convergent results of the segmentation, indicating good quality and stability of the obtained descriptive models. The advantage of these methods is a clear way of presenting the results. Descriptive models can be efficient tools for fast comparative evaluation of the investigated extracts in terms of their polyphenols fraction content and antioxidant properties Analysis of the relationship between antioxidant activity and total polyphenol content The relationship between the total content of polyphenols in the studied extracts and their ability to inactivate free radicals was assessed. For this purpose, a simple linear regression analysis was performed. A statistically 705

7 Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts Figure 4. Projection of the investigated extracts on the space of the first three principal components. significant positive correlation was found (p < 0.01) between the content of polyphenols and antioxidant activity expressed as -log(ic 50 DPPH ) and -log(ic ) 50 ABTS (r = and r = 0.990, respectively). As can be seen in Fig. 5, the extracts with a higher content of polyphenols show the stronger antiradical activity against both DPPH radicals and ABTS + cation radicals. Two dietary supplements (DS2, DS4) indicate the highest -log(ic 50 ) values. These herbal products contain highly concentrated extracts from the leaves of Camellia sinensis. The dietary supplement containing the standardized extract of Camellia sinensis (DS5) also indicates a high -log(ic 50 ) value and thus a high antioxidant activity. The results of the simple linear regression analysis suggest that radical scavenging ability depends on the amount of total polyphenolic compounds in Camellia sinensis extract. The extracts rich in polyphenols show much better DPPH and ABTS + scavenging ability than those containing less phenolic compounds. Furthermore, the analysis of IC 50 values shows that the values obtained in ABTS + scavenging activity assay are lower than those obtained in DPPH radical scavenging activity assay. This is due to the fact that ABTS + cation radicals are much more reactive than DPPH radicals and, unlike the reactions with DPPH radicals, which involve a hydrogen atom transfer, the reactions with ABTS + radicals involve only an electron transfer process [31] Predictive model The tree regression algorithm was applied to construct the calibration model for predicting the antioxidant activity (expressed as the IC 50 value) of investigated Camellia sinensis extracts on the basis of their fingerprint chromatograms. In order to correct the retention time shifts, the peak alignment of the chromatograms was performed by using a correlation optimized warping function with linear interpolation [32]. The advantage of a nonparametric tree regression algorithm is lack of initial assumptions about the data distributions. The input data set was divided into two subsets (testing and training), so that a reliable predictive model was obtained. The training subset contained 75% of all cases, whereas the other cases formed the testing subset, which was needed to test a predictive ability of the calibration model. The predictive model included only predictors of a high impact on the predicted value of the dependent variable, the variables of poor predictive values were excluded from the model construction. The results of predictive modeling are presented in the form of a tree graph (Fig. 6). A binary tree (only two child nodes were formed in each division) with a set of decision rules was obtained. These decision rules describe and explain the effect of the independent variables (chromatographic data) on the antioxidant activity of Camellia sinensis extract (expressed as IC 50 ABTS value). Due to the known mechanism of the tree construction and clear sequence of decision rules, the 706

8 J. Ronowicz, B. Kupcewicz, E. Budzisz Figure 5. Scatterplot of -log(ic 50 DPPH ) and -log(ic ) values against total polyphenol content. 50 ABTS Figure 6. The tree graph for IC 50 ABTS value. 707

9 Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts Table 3. Summary table collecting the samples with experimental and predicted data for training and testing subsets. Code of herbal Product Subset Experimental value Predicted value Standard error Residual GT1 testing GT1 testing GT1 testing GT2 training GT2 training GT2 training GT3 training GT3 training GT3 training GT4 testing GT4 testing GT4 testing GT5 training GT5 training GT5 training GT6 training GT6 training GT6 training DS1 training DS1 training DS2 training DS2 training DS2 training DS3 training DS3 training DS3 training DS4 testing DS4 testing DS4 testing DS5 training DS5 training DS5 training DS6 training DS6 training DS6 training interpretation of the obtained model is possible. The analysis of the regression tree indicates that a chemical compound eluted from the chromatographic column with retention time t R = min has the highest predictive value. If the signal recorded by a spectrophotometric detector in this time is larger than a.u., the extract indicates the highest antioxidant properties (a low value of IC 50 ABTS = 9.28 μg ml -1 ). The lowest antioxidant activity (a high value of IC 50 ABTS ) shows the extracts for which the signal recorded by the detector at t R = min is less than or equal to a.u. and simultaneously the signal recorded at t R = 708

10 J. Ronowicz, B. Kupcewicz, E. Budzisz min is less than or equal to a.u. The analysis of decision rules indicates that a chemical compound eluted from the chromatographic column with retention time t R = min also determines the antioxidant activity of the extracts. The model performance was estimated as the root mean squared error for the training and testing subsets: 2.10 and 5.83, respectively. The model fitting R 2 was and the predictive ability Q 2 was However, in order to consider a model as final and fully complete, more cases should be included to the training and testing subset. A summary table (Table 3) collecting the samples with experimental and predicted data for training and testing subsets is given above. Application of regression trees to predict the IC 50 value based on the fingerprint chromatogram could enable quick verification of the quality of the extract in terms of its antioxidant properties. The advantages of the tree regression model are simplicity and good interpretability. The analysis of the obtained decision rules indicates that the chemical compound eluted from the chromatographic column with retention time t R = min is mainly responsible for the antioxidant properties of Camellia sinensis extract. Two chemical compounds eluted from the chromatographic column with retention time t R = min and t R = min also determine the antioxidant activity of Camellia sinensis extract. Application of this chemometric technique allows us to obtain the information about the antioxidant activity from the fingerprint chromatogram. 4. Conclusions Chemometric techniques allowed for efficient comparison of investigated Camellia sinensis extracts in terms of their antioxidant properties. The extracts exhibited varied antioxidant activity. Four dietary supplements showed the highest activity against both DPPH radicals and ABTS + cation radicals. It has been shown that the ability of the Camellia sinensis extracts to inactivate free radicals is closely related with the content of polyphenolic compounds. The use of multivariate calibration techniques allowed for prediction of antioxidant activity of the extract on the basis of the obtained fingerprint chromatogram. This activity may be one of the measures of the quality of an herbal product and thus its therapeutic efficacy in diseases associated with the oxidative disorders in cells and tissues (such as neurodegenerative, cardiovascular and atherosclerotic disorders). Acknowledgements The authors gratefully acknowledge the financial support (DS-UPB 833) from Collegium Medicum, Nicolaus Copernicus University in Toruń (Poland). References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] L. Standley, P. Winterton, J.L. Marnewick, W.C.A. Gelderblom, E. Joubert, T.J. Britz, J. Agric. Food Chem. 49, 114 (2001) T. Yokozawa, T. Nakagawa, K. Kitani, J. Agric. Food Chem. 44, 3549 (2002) N. Zaveri, Life Sci. 78, 2073 (2006) C.S. Yang, P. Maliakal, X. Meng, Annu. Rev. Pharmacol. Toxicol. 42, 25 (2002) P. Carloni, L. Tiano, L. Padella, T. Bacchetti, Ch. Customu, A. Kay, E. Damiani, Food Res. Int. 53, 900 (2013) S. Gorjanovic, D. Komes, F.T. Pastor, A. Bels c ak-cvitanovic, L. Pezo, I. Hec imovic, D. Suz njevic, J. Agric. Food Chem. 60, 9573 (2012) Y. Liang, P. Xie, F. Chau, J. Sep. Sci. 33, 410 (2010) H.A. Gad, Sh.H. El-Ahmady, M.I. Abou-Shoerb, M.M. Al-Azizi, Phytochem. Anal. 24, 1 (2013) B. Xie, T. Gong, M. Tang, D. Mi, X. Zhang, J. Liu, Z. Zhang, J. Pharm. Biomed. Anal. 48, 1261 (2008) Ch. Chen, J. Yuan, X. Li, Z. Shen, D. Yu, J. Zhu, [11] [12] [13] [14] [15] [16] [17] F. Zeng, Chem. Biol. Drug. Des. 81, 688 (2013) M. Dumarey, A.M. van Nederkassel, E. Deconinck, Y. Vander Heyden, J. Chromatogr. A 1192, 81 (2008) M. Dumarey, I. Smets, Y. Vander Heyden, J. Chromatogr. B 878, 2733 (2010) A.M. van Nederkassel, M. Daszykowski, D.L. Massart, Y. Vander Heyden, J. Chromatogr. A 1096, 177 (2005) M. Daszykowski, Y. Vander Heyden, B. Walczak, J. Chromatogr. A 1176, 12 (2007) C. Chen, S.X. Li, S.M. Wang, S.W. Liang, J. Pharm. Biomed. Anal. 56, 443 (2011) C. Tistaert, B. Dejaegher, G. Chataigné, C. Van Minh, J. Quetin-Leclercq, Y. Vander Heyden, Talanta 83, 1198 (2011) C. Tistaert, B. Dejaegher, N. Nguyen Hoai, G. Chataigné, C. Rivic re, V. Nguyen Thi Hong, M. Chau Van, J. Quetin-Leclerq, Y. Vander Heyden, Anal. Chim. Acta 649, 24 (2009) 709

11 Implementation of chemometric techniques for evaluation of antioxidant properties of Camellia sinensis extracts [18] C. Tistaert, B. Dejaegher, G. Chataigné, C. Rivière, N. Nguyen Hoai, M. Chau Van, J. Quetin-Leclercq, Y. Vander Heyden, Anal. Chim. Acta 721, 35 (2012) [19] A. Buciński, H. Zieliński, H. Kozłowska, Trends Food Sci. Tech. 15, 161 (2004) [20] A. Djeridane, M. Yousfi, B. Nadjemi, D. Boutassouna, P. Stocker, N. Vidal, Food Chem. 97, 654 (2006) [21] O. Dóka, D. Bicanic, Anal. Chem. 74, 2157 (2002) [22] I. Gűlçin, Arch. Toxicol. 86, 345 (2012) [23] P. Molyneux, Songklanakarin J. Sci. Technol. 26, 211 (2004) [24] F. Pellati, S. Benvenuti, L. Magro, M. Melegari, F. Soragni, J. Pharm. Biomed. Anal. 35, 289 (2004) [25] Y. Cai, Q Luo, M. Sun, H. Corke, Life Sci. 74, 2157 (2004) [26] T. Hill, P. Lewicki, Statistics: Methods and Applications (StatSoft Inc., Tulsa, Oklahoma, 2006) [27] [28] [29] [30] [31] [32] T. Hastie, R. Tibshirani, J. Friedman, The elements of statistical learning: Data mining, inference, and prediction, 2nd edition (Springer-Verlag, New York, 2009) R. Nisbet, J. Elder, G. Miner, Handbook of Statistical Analysis and Data Mining Applications (Elsevier, Oxford, 2009) M. Daszykowski, B. Walczak, D.L. Massart, Chemometr. Intell. Lab. 56, 83 (2001) A.J. Chartlon, M.S. Wrobel, I. Stanimirowa, M. Daszykowski, H.H. Grundy, B. Walczak, Eur. Food Res. Technol. 231, 733 (2010) S. Kaviarasan, G.H. Naik, R. Gangabhagirathi, C.V. Anuradha, K.I. Priyadarsini, Food Chem. 103, 31 (2007) G. Tomasi, F. van den Berg, C. Andersson, J. Chemometr. 18, 231 (2004) 710