Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
Rapid determination of major bioactive isoflavonoid compounds during the extraction process of kudzu (Pueraria lobata) by near-infrared transmission spectroscopy Pei Wang a,b,1, Hui Zhang c,1, Hailong Yang a, Lei Nie a, Hengchang Zang a,⇑ a b c
National Glycoengineering Research Center and School of Pharmaceutical Science, Shandong University, N0. 44 West Wenhua Road, Jinan 250012, China School of Pharmacy, Shenyang Pharmaceutical University, Shenyang 110016, China Beijing Kaiyuan Shengshi Science and Technology Development Co., LTD., Jinan 250012, China
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
NIRS was used to determine puerarin,
daidzin, daidzein and total isoflavonoid (TIF) during the extraction process of kudzu. Different variable selection methods were employed for variable selection and model optimization. 2 Excellent R (0.985–0.997), RMSEC (0.016–0.095) and RMSEP (0.016– 0.145) values obtained. Possibility of implementation in kudzu extracts production.
a r t i c l e
i n f o
Article history: Received 27 July 2014 Received in revised form 27 August 2014 Accepted 1 September 2014 Available online 8 September 2014 Keywords: Near-infrared spectroscopy Partial least squares regression High performance liquid chromatography Kudzu
a b s t r a c t Near-infrared (NIR) spectroscopy has been developed into an indispensable tool for both academic research and industrial quality control in a wide field of applications. The feasibility of NIR spectroscopy to monitor the concentration of puerarin, daidzin, daidzein and total isoflavonoid (TIF) during the extraction process of kudzu (Pueraria lobata) was verified in this work. NIR spectra were collected in transmission mode and pretreated with smoothing and derivative. Partial least square regression (PLSR) was used to establish calibration models. Three different variable selection methods, including correlation coefficient method, interval partial least squares (iPLS), and successive projections algorithm (SPA) were performed and compared with models based on all of the variables. The results showed that the approach was very efficient and environmentally friendly for rapid determination of the four quality indices (QIs) in the kudzu extraction process. This method established may have the potential to be used as a process analytical technological (PAT) tool in the future. Ó 2014 Elsevier B.V. All rights reserved.
Introduction
⇑ Corresponding author. Tel./fax: +86 531 88380268. E-mail address: [email protected] (H. Zang). These authors contributed equally to this work and should be regarded as co-first authors. 1
http://dx.doi.org/10.1016/j.saa.2014.09.002 1386-1425/Ó 2014 Elsevier B.V. All rights reserved.
Kudzu, the dried root of Pueraria lobata (Wild.) Ohwi, is one of the earliest and most important edible herbs used in oriental medicine. It has been widely used in eastern Asia for the treatment of deafness, acute dysentery, diarrhea and cardiovascular diseases [1]. Modern studies of kudzu have shown that its extract exhibits
1404
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
antioxidant, radical scavenging, anti-cancer, antidepressant and neuro-protective activities [2–6]. Kudzu extract is rich in isoflavonoids, which are proposed to be responsible for most of biological activities [7]. Several isoflavonoids, such as puerarin, daidzin and daidzein, were isolated from kudzu and their therapeutic effects have been reported [5,8–10]. Thus, the isoflavonoid content is an important indicator for the quality of kudzu samples, as well as kudzu-containing preparations. Extraction is not only one of the most important manufacturing units, but also the most initial process for various kudzu extracts and kudzu-containing preparations. The efficiency of different extraction methods such as traditional, pressurized solvent extraction and ultrasonic techniques have been investigated [11]. Among the extraction methods, ultrasonic technique would be more efficient than the others for the manufacturing of kudzu extract [11]. However, in large-scale manufacturing process, the conventional extraction method is still the mainstream approach, and heat reflux extraction (HRE) is the most widely used technique for the extraction of puerarin from kudzu [12]. Usually, the extraction process ends when the stipulated extraction time is reached. The influence of temperature, pressure and fluctuations of crude drug quality (e.g. Kudzu materials) on extracts is generally ignored [13]. To improve process efficiency and guarantee final product quality, reliable process analytical technology (PAT) should be emphasized in the manufacturing process of traditional Chinese medicine (TCM) and dietary supplements, especially in the extraction unit. Near-infrared (NIR) spectroscopy, associated with chemometrics, is becoming a powerful tool with lots of advantages such as fast acquisition, non-invasive, non-destructive, minimization of sample preparation, and capability to handle and control a great number of industrial/technological variables which must be optimized in manufacturing processes [14–16]. NIR spectroscopy has also shown great power and gained wide acceptance in TCM [17]. The most well-known uses of NIR spectroscopy in the manufacturing process of TCM include qualitative discrimination and quantitative determination of herb materials, intermediates and products [18–21], separation monitoring [22], end point assessment of extraction process [13,23], and alcohol precipitation monitoring [24]. Obviously, chemical information in herbal materials and herbal products can be reflected in NIR spectra, and process monitoring using NIR spectroscopy technology can play an important role in process understanding. The objective of this study is to investigate the feasibility and application of NIRS in determining puerarin, daidzin, daidzein and total isoflavonoid (TIF) concentrations during the extraction process of kudzu. NIR spectroscopy calibration models were established using partial least squares (PLS) based on the NIR spectral and reference data collected from the kudzu extraction process. Three variable selection methods including correlation coefficient method, interval partial least squares (iPLS) and successive projections algorithm (SPA) were employed for variable selection and model optimization. The performance of the models were evaluated using the correlation coefficients and prediction errors in both calibration and validation steps. To our best knowledge, this research was the first to report the NIR spectroscopy method for the determination of isoflavonoids during kudzu extraction process, and could also be used as a reference to achieve similar purposes of other herb materials.
Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). Ethanol (70%, v/v) used as the extraction solvent was medicinal grade. HPLC grade acetonitrile was obtained from Fisher Scientific (Pittsburgh, PA). Water was purified by Millipore water purification device (Millipore, USA). All the other reagents were of analytical grade. Sample preparation Kudzu (8 kg) was extracted with 8-fold 70% ethanol at 85 ± 2 °C and repeated twice. The first extraction process lasted 90 min and the second lasted 60 min. Samples collected at different times were centrifuged at 8000 rpm for 10 min. The supernatant was analyzed by NIR, HPLC and UV, respectively. In order to obtain similar prediction accuracy, it was necessary to ensure uniform distribution of samples with high or low concentrations [13]. In the first extraction process, extract samples were collected after 20 min and at 5 min intervals over the subsequent 70 min. In the second extraction process, extract samples were collected after 18 min and at 7 min intervals over the subsequent 42 min. In this study, four extraction batches were carried out and 88 samples were collected. Spectra acquisition The NIR spectra were performed by an Antaris II Fourier transform near-infrared spectrophotometer (Thermo Fisher Scientific, USA). Transmission spectra of samples were collected from 10,000 to 4000 cm 1 at every 4 cm 1 path intervals. Each spectrum was obtained by averaging 32 scans and the gain was auto-optimized to increase the signal to noise ratio. To avoid error from the outer environment, all samples were equilibrated to room temperature (25 °C) prior to NIR analysis. And the humidity was also kept at ambient level in the laboratory. Determination of reference values Determination of reference values was performed immediately after the NIR measurements. After appropriate dilution with 70% ethanol, the TIF was determined using UV spectroscopy reported by Qingxin Kong [25], and the validated reversed-phase HPLC assay [26] with a slight modification was applied to the quantitative determination of puerarin, daidzin and daidzein. Chromatographic analysis was performed on an Agilent 1260 HPLC system (Agilent Corp., Santa Clara, CA, USA). A Phenomenex C18 column (250 mm 4.6 mm i.d. 5 lm) was employed. The mobile phase was composed of acetonitrile (A) and 0.05% phosphoric acid in water (B). The gradient elution had the following profile: 8–12% (A) in 0–21 min, 12–17% (A) in 21–31 min, 17–38% (A) in 31– 55 min, 38–90% (A) in 55–65 min and 90% (A) in 65–75 min, at a flow rate of 1.0 mL/min. The wavelength of UV detector was set at 278 nm and the injection volume was 10 lL. Data processing All the computations, including data collecting and loading, division of the calibration and validation set, spectral pre-processing, variable selection, chemometric regression model construction and validation were performed using MATLABÒ version 7.10 environment (Math-Works, Natick, USA) with the PLS-toolbox version 4.0 (Eigenvector Research, Inc., Wenatchee, WA) and TQ analyst software package (Version 8.0, Thermo Scientific, Madison, WI, USA).
Materials and methods Results and discussion Plant materials and reagents Reference values analysis Kudzu materials were supplied by a TCM pharmaceutical factory (Shandong Wohua Pharmaceututial Co., Ltd., Weifang, China). Puerarin, daidzin and daidzein were purchased from the National
Using the HPLC and UV spectrophotometric methods described in Section ‘Determination of reference values’, all 88 samples were
1405
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
set while the remaining samples were split into two groups, calibration and validation sets, by K–S algorithm. In this research, 60 samples were placed into the calibration set, and the remaining 28 samples were placed into the validation set. The statistical values are listed in Table 2. After division by K–S algorithm, the content values of isoflavonoids in the calibration and the validation set can cover a wide enough range, which is helpful to develop robust models.
Fig. 1. Characteristic chromatograms of the mixed standards (A) and kudzu solution sample (B). (1. Puerarin, 2. Daidzin, 3. Daidzein.)
analyzed. Representative HPLC chromatograms of the standard solution (three isoflavonoid components) and the extract solution of kudzu are shown in Fig. 1. Baseline separation for all three analytes can be achieved. Before the sample testing, both the HPLC and UV spectrophotometric methods were validated. The main methodology parameters and calibration curves of the reference HPLC and UV method are listed in Table 1. Analysis of raw NIR spectra Fig. 2 shows the NIR spectra of 88 kudzu extract samples. There are no apparent differences in the raw NIR spectra of samples due to high-degree band overlapping. However, the usable spectral regions are necessary to be identified prior to developing the calibration models. The NIR spectra exhibit intense absorption bands around 7000 cm 1 from the first O–H overtone and around 5160 cm 1 from the combination of stretching and deformation of the O–H group in water [27]. These two bands, which are typical characteristics of the NIR spectrum of an aqueous alcoholic solution [28], hold a significant influence on determination of the active constituents with near infrared spectroscopic technique because the intense bands could mask any other bands present in these spectral ranges [28,29]. The 7500–10,000 cm 1 region characterized by low intensity and low signal-to-noise ratio is assigned to the second and third overtones [27]. The 4500– 5000 cm 1 region corresponds to combinations of fingerprint absorptions with the C–H, O–H and N–H stretching modes [30]. The 5450–6100 cm 1 region, which is related to the first overtones of the C–H stretching mode [27], is useful for quantitative NIR models, although this spectral region is always immersed in the intense water band [28]. Quantitative calibration models Division of calibration and validation set The samples with maximum and minimum concentrations of puerarin, daidzin, daidzein and TIF were included in the calibration
Optimization of calibration model In this study, for the puerarin, daidzin, daidzein and TIF models, the ‘‘fingerprint region’’ (4500–5000 cm 1), and the ‘‘water regions’’ (5000–5450 cm 1 and 6600–7500 cm 1) were excluded. The spectral region actually employed was selected from 5450 to 6600 cm 1 for daidzin and daidzein; and for puerarin and TIF which have relatively high concentrations (the concentration of puerarin higher than 0.4 mg/ml and TIF higher than 0.92 mg/ml), the region of 7500–10,000 cm 1 was also employed. To confirm the correctness of the band selection, three different variable selection methods including correlation coefficient method, iPLS and SPA were performed. Using the TIF as an example, as shown in Fig. 3, the variables selected by forward iPLS are mostly distributed in the selected region. Similar cases can also be observed for the other three quality indices (QIs). In addition, different spectral pretreatments were investigated in an effort to optimize calibration performance. As for the transmitted or transflective spectra, the most frequently used pretreated method is derivative (D) calculation, including 1st D, 2nd D, or 3rd D. To avoid noise magnification caused by derivative calculation, spectra are first smoothed using the Savitzky–Golay filter (S–G) algorithm, which is a moving window method [21]. In the present study, different spectral pretreated methods were compared and selected for each calibration model. The decision on which pre-processed method to adopt was based on the cross-validation results. The selected wave-number intervals and optimized spectral pretreatment methods for the calibration models are listed in Table 3. Determination of optimal latent variable numbers The PLSR reduces the dimensionality of spectral data through the calculation of the latent variables (LVs), which explain the maximum amount of variability in the data. Every PLS calibration model has an optimal number of LVs, which are considered as the best fit for the original spectral variability. LVs lower or greater than the optimal one introduced in the model bring about the problem of ‘‘under-fitting’’ or ‘‘over-fitting’’, and both of which will decrease predictability. Cross-validation via the venetian blinds algorithm (root-mean-squared error of cross-validation, RMSECV) was used for seeking the optimal LVs. The most suitable LVs for the models of four QIs are also shown in Table 3. Calibration and validation of the quantitative models The performance of models was assessed by coefficient of calibration (Rc), coefficient of validation (Rp), root mean square error of calibration (RMSEC), root mean square errors of cross-validation (RMSECV) and root meat square errors of prediction (RMSEP). A
Table 1 Method parameters and linear calibration functions of the HPLC and UV spectrophotometric method.
a b
Compound
tR (min)
Linearity range (lg/mL)
Calibration curve
Puerarina Daidzina Daidzeina TIFb
17.10 23.06 35.41 –
2.50–500.0 0.60–120.0 0.50–40.0 0.16–11.2
Y = 3.0 10 Y = 3.0 10 Y = 2.0 10 Y = 4.3 10
Determinated by HPLC. Determinated by UV spectrophotometric.
4 4 4 2
X + 2.7 10 X + 1.3 10 X + 6.0 10 X + 1.4 10
3 3 4 3
R
Repeatability (RSD%, n = 6)
Recovery (%, n = 6)
1.0000 1.0000 1.0000 0.9999
1.15 1.27 0.93 1.33
101.2 98.6 99.7 100.5
1406
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
Fig. 2. Raw spectra of kudzu solution in the water–ethanol extraction process.
Table 2 Statistical values of the isoflavonoids content in calibration and validation sets. Compound
Puerarin Daidzin Daidzein TIF
Total sets (mg/ml)
Calibration sets (mg/ml)
Validation sets (mg/ml)
Average values
Variation ranges
Average values
Variation ranges
Average values
Variation ranges
1.98 0.31 0.57 4.64
0.40–3.42 0.05–0.63 0.09–0.94 0.92–8.58
1.99 0.29 0.55 4.60
0.40–3.42 0.05–0.63 0.09–0.94 0.92–8.58
1.95 0.34 0.61 4.74
0.58–3.00 0.12–0.53 0.18–1.02 1.60–7.40
Fig. 3. The variable regions selected by forward iPLS.
Table 3 The most suitable conditions, and performance parameters of the calibration models established via PLSR. Spectral pretreatment method
Optimum no. of LVs
R2
RMSEC (mg/mL)
RMSECV (mg/mL)
RMSEP (mg/mL)
8649–8458, 8070–7504, 6527–5951
Autoscale,1st D, SG (15, 2)
6
0.996
0.045
0.058
0.063
Daidzin
6099–5499
Autoscale,1st D, SG (15, 2)
6
0.989
0.016
0.019
0.016
Daidzein
6099–5399
Autoscale,1st D, SG (15, 2)
5
0.985
0.032
0.038
0.034
TIF
9420–9229, 6144–5951, 5563–5372
Autoscale,1st D, SG (15, 2)
7
0.997
0.095
0.127
0.145
Compound
Wavenumber interval (cm
Puerarin
1
)
good NIR calibration model should have low RMSEC and RMSEP values, high correlation coefficient (R), but small differences between RMSECV and RMSEP. According to the procedures mentioned above, the calibration models of the four QIs were established. As shown in Table 3, the
RMSEC and R for calibration of TIF are 0.095 and 0.997, respectively, and the RMSEP and RMSECV for the validation are 0.145 and 0.127, respectively. The performance of parameters of the other three QIs models are listed in Table 3. The results show that the established models give satisfactory fitting results and predic-
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
1407
Fig. 4. The correlation charts of the four calibration models ((A) TIFs; (B) Puerarin; (C) Daidzin; (D) Daidzein).
tive ability, and can be applied to monitoring the concentration of the four QIs in the water–ethanol extraction process of kudzu. Fig. 4 presents the relevance plots between reference values (xaxis) and the PLS prediction values (y-axis) for all the four QIs. Prediction accuracy is satisfactory.
Conclusion In this study, a NIR spectroscopy method is explored and established for the simultaneous determination of puerarin, daidzin, daidzein and TIF during the water–ethanol extraction process of kudzu. By means of PLSR, robust NIR models were built between transmissive spectra and reference values with high correlations and low error of predictions. Compared with the HPLC and UV methods, the presented method based on NIRS can significantly save manpower and time. The established calibration models are applying to monitor the extraction process of kudzu by our company partner, and the successful implementation can also help other factories to achieve similar purposes. Acknowledgements We are grateful for the financial support of the Major Projects of Independent Innovation Achievements of Shandong Province (No. 2010ZDZX1A0406) and Shandong Province Natural Science Foundation (No. ZR2011HM080). The authors would like to acknowledge the collaborators from Wohua Pharmaceutical Co., Ltd., for supporting this work and for permission to publish the above results. References [1] K.H. Wong, G.Q. Li, K.M. Li, V. Razmovski-Naumovski, K. Chan, Kudzu root: traditional uses and potential medicinal benefits in diabetes and cardiovascular diseases, J. Ethnopharmacol. 134 (2011) 584–607. [2] B. Yan, D.-Y. Wang, D.-M. Xing, Y. Ding, R.-F. Wang, F. Lei, L.-J. Du, The antidepressant effect of ethanol extract of radix puerariae in mice exposed to cerebral ischemia reperfusion, Pharmacol. Biochem. Behav. 78 (2004) 319– 325. [3] H. Kim, Neuroprotective herbs for stroke therapy in traditional eastern medicine, Neurol. Res. 27 (2005) 287–301. [4] Y. Wenli, Z. Yaping, S. Bo, The radical scavenging activities of radix puerariae isoflavonoids: a chemiluminescence study, Food Chem. 86 (2004) 525–529. [5] W.M. Keung, B.L. Vallee, Kudzu root: an ancient Chinese source of modern antidipsotropic agents, Phytochemistry 47 (1998) 499–506. [6] H. Yao, B. Wu, Y. Cheng, H. Qu, High throughput chemiluminescence platform for evaluating antioxidative activity of total flavonoid glycosides from plant extracts, Food Chem. 115 (2009) 380–386.
[7] Z. Zhang, T.N. Lam, Z. Zuo, Radix puerariae: an overview of its chemistry, pharmacology, pharmacokinetics, and clinical use, J. Clin. Pharmacol. 53 (2013) 787–811. [8] L. Coward, N.C. Barnes, K.D.R. Setchell, S. Barnes, Genistein, daidzein, and their beta-glycoside conjugates: antitumor isoflavones in soybean foods from American and Asian diets, J. Agric. Food Chem. 41 (1993) 1961–1967. [9] D.K.Y. Yeung, S.W.S. Leung, Y.C. Xu, P.M. Vanhoutte, R.Y.K. Man, Puerarin, an isoflavonoid derived from radix puerariae, potentiates endotheliumindependent relaxation via the cyclic AMP pathway in porcine coronary artery, Eur. J. Pharmacol. 552 (2006) 105–111. [10] P. Van Hung, N. Morita, Chemical compositions, fine structure and physicochemical properties of kudzu (Pueraria lobata) starches from different regions, Food Chem. 105 (2007) 749–755. [11] M.-H. Lee, C.-C. Lin, Comparison of techniques for extraction of isoflavones from the root of radix puerariae: ultrasonic and pressurized solvent extractions, Food Chem. 105 (2007) 223–228. [12] J.P. Fan, J. Cao, X.H. Zhang, J.Z. Huang, T. Kong, S. Tong, Z.Y. Tian, Y.L. Xie, R. Xu, J.H. Zhu, Optimization of ionic liquid based ultrasonic assisted extraction of puerarin from radix puerariae lobatae by response surface methodology, Food Chem. 135 (2012) 2299–2306. [13] Y. Wu, Y. Jin, H. Ding, L. Luan, Y. Chen, X. Liu, In-line monitoring of extraction process of scutellarein from Erigeron breviscapus (vant.) Hand-Mazz based on qualitative and quantitative uses of near-infrared spectroscopy, Spectrochim. Acta Part A Mol. Biomol. Spectrosc. 79 (2011) 934–939. [14] J.J. Workman, Review of process and non-invasive near-infrared and infrared spectroscopy: 1993–1999, Appl. Spectrosc. Rev. 34 (1999) 1–89. [15] G. Reich, Near-infrared spectroscopy and imaging: basic principles and pharmaceutical applications, Adv. Drug Deliv. Rev. 57 (2005) 1109–1143. [16] Y. Roggo, P. Chalus, L. Maurer, C. Lema-Martinez, A. Edmond, N. Jent, A review of near infrared spectroscopy and chemometrics in pharmaceutical technologies, J. Pharm. Biomed. Anal. 44 (2007) 683–700. [17] A. Rohman, A. Nugroho, E. Lukitaningsih, Sudjadi, Application of vibrational spectroscopy in combination with chemometrics techniques for authentication of herbal medicine, Appl. Spectrosc. Rev. 49 (2014) 603–613. [18] J. Lu, B. Xiang, H. Liu, S. Xiang, S. Xie, H. Deng, Application of two-dimensional near-infrared correlation spectroscopy to the discrimination of Chinese herbal medicine of different geographic regions, Spectrochim. Acta Part A, Mol. Biomol. Spectrosc. 69 (2008) 580–586. [19] Y.A. Woo, H.J. Kim, J.H. Cho, H. Chung, Discrimination of herbal medicines according to geographical origin with near infrared reflectance spectroscopy and pattern recognition techniques, J. Pharm. Biomed. Anal. 21 (1999) 407– 413. [20] Y. Feng, D. Lei, C. Hu, Rapid identification of illegal synthetic adulterants in herbal anti-diabetic medicines using near infrared spectroscopy, Spectrochim. Acta Part A, Mol. Biomol. Spectrosc. 125 (2014) 363–374. [21] W. Li, Z. Cheng, Y. Wang, H. Qu, A study on the use of near-infrared spectroscopy for the rapid quantification of major compounds in Tanreqing injection, Spectrochim. Acta Part A, Mol. Biomol. Spectrosc. 101 (2013) 1–7. [22] H. Liu, X. Zhao, T. Qi, Y.P. Qi, G.R. Fan, Establishment of the model for online monitoring of the column separation and purification process by near-infrared spectroscopy and determination of total ginsenosides in Folium Ginseng, Guang pu xue yu guang pu fen xi = Guang pu 33 (2013) 3226–3230. [23] Z. Wu, C. Sui, B. Xu, L. Ai, Q. Ma, X. Shi, Y. Qiao, Multivariate detection limits of on-line NIR model for extraction process of chlorogenic acid from Lonicera japonica, J. Pharm. Biomed. Anal. 77 (2013) 16–20. [24] Z. Wu, B. Xu, M. Du, C. Sui, X. Shi, Y. Qiao, Validation of a NIR quantification method for the determination of chlorogenic acid in Lonicera japonica solution in ethanol precipitation process, J. Pharm. Biomed. Anal. 62 (2012) 1–6. [25] K. Qingxin, Process study of extracting total isoflavones from radix puerariae by microwave-assisted extraction, J. Anhui Agric. Sci. 35 (2007) 10835.
1408
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
[26] P. Wang, L. Li, H. Yang, S. Cheng, Y. Zeng, L. Nie, H. Zang, Chromatographic fingerprinting and quantitative analysis for the quality evaluation of Xinkeshu tablet, J. Pharm. Anal. 2 (2012) 422–430. [27] N. Gierlinger, M. Schwanninger, R. Wimmer, Characteristics and classification of Fourier-transform near infrared spectra of the heartwood of different larch species (Larix sp.), J. Near Infrared Spectrosc. 12 (2004) 113–119. [28] Y. Jin, Z. Wu, X. Liu, Y. Wu, Near infrared spectroscopy in combination with chemometrics as a process analytical technology (PAT) tool for on-line
quantitative monitoring of alcohol precipitation, J. Pharm. Biomed. Anal. 77 (2013) 32–39. [29] Q. Dong, H. Zang, A. Liu, G. Yang, C. Sun, L. Sui, P. Wang, L. Li, Determination of molecular weight of hyaluronic acid by near-infrared spectroscopy, J. Pharm. Biomed. Anal. 53 (2010) 274–278. [30] K.-Z. Liu, M. Shi, A. Man, T.C. Dembinski, R.A. Shaw, Quantitative determination of serum LDL cholesterol by near-infrared spectroscopy, Vib. Spectrosc. 38 (2005) 203–208.
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
Rapid determination of major bioactive isoflavonoid compounds during the extraction process of kudzu (Pueraria lobata) by near-infrared transmission spectroscopy Pei Wang a,b,1, Hui Zhang c,1, Hailong Yang a, Lei Nie a, Hengchang Zang a,⇑ a b c
National Glycoengineering Research Center and School of Pharmaceutical Science, Shandong University, N0. 44 West Wenhua Road, Jinan 250012, China School of Pharmacy, Shenyang Pharmaceutical University, Shenyang 110016, China Beijing Kaiyuan Shengshi Science and Technology Development Co., LTD., Jinan 250012, China
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
NIRS was used to determine puerarin,
daidzin, daidzein and total isoflavonoid (TIF) during the extraction process of kudzu. Different variable selection methods were employed for variable selection and model optimization. 2 Excellent R (0.985–0.997), RMSEC (0.016–0.095) and RMSEP (0.016– 0.145) values obtained. Possibility of implementation in kudzu extracts production.
a r t i c l e
i n f o
Article history: Received 27 July 2014 Received in revised form 27 August 2014 Accepted 1 September 2014 Available online 8 September 2014 Keywords: Near-infrared spectroscopy Partial least squares regression High performance liquid chromatography Kudzu
a b s t r a c t Near-infrared (NIR) spectroscopy has been developed into an indispensable tool for both academic research and industrial quality control in a wide field of applications. The feasibility of NIR spectroscopy to monitor the concentration of puerarin, daidzin, daidzein and total isoflavonoid (TIF) during the extraction process of kudzu (Pueraria lobata) was verified in this work. NIR spectra were collected in transmission mode and pretreated with smoothing and derivative. Partial least square regression (PLSR) was used to establish calibration models. Three different variable selection methods, including correlation coefficient method, interval partial least squares (iPLS), and successive projections algorithm (SPA) were performed and compared with models based on all of the variables. The results showed that the approach was very efficient and environmentally friendly for rapid determination of the four quality indices (QIs) in the kudzu extraction process. This method established may have the potential to be used as a process analytical technological (PAT) tool in the future. Ó 2014 Elsevier B.V. All rights reserved.
Introduction
⇑ Corresponding author. Tel./fax: +86 531 88380268. E-mail address: [email protected] (H. Zang). These authors contributed equally to this work and should be regarded as co-first authors. 1
http://dx.doi.org/10.1016/j.saa.2014.09.002 1386-1425/Ó 2014 Elsevier B.V. All rights reserved.
Kudzu, the dried root of Pueraria lobata (Wild.) Ohwi, is one of the earliest and most important edible herbs used in oriental medicine. It has been widely used in eastern Asia for the treatment of deafness, acute dysentery, diarrhea and cardiovascular diseases [1]. Modern studies of kudzu have shown that its extract exhibits
1404
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
antioxidant, radical scavenging, anti-cancer, antidepressant and neuro-protective activities [2–6]. Kudzu extract is rich in isoflavonoids, which are proposed to be responsible for most of biological activities [7]. Several isoflavonoids, such as puerarin, daidzin and daidzein, were isolated from kudzu and their therapeutic effects have been reported [5,8–10]. Thus, the isoflavonoid content is an important indicator for the quality of kudzu samples, as well as kudzu-containing preparations. Extraction is not only one of the most important manufacturing units, but also the most initial process for various kudzu extracts and kudzu-containing preparations. The efficiency of different extraction methods such as traditional, pressurized solvent extraction and ultrasonic techniques have been investigated [11]. Among the extraction methods, ultrasonic technique would be more efficient than the others for the manufacturing of kudzu extract [11]. However, in large-scale manufacturing process, the conventional extraction method is still the mainstream approach, and heat reflux extraction (HRE) is the most widely used technique for the extraction of puerarin from kudzu [12]. Usually, the extraction process ends when the stipulated extraction time is reached. The influence of temperature, pressure and fluctuations of crude drug quality (e.g. Kudzu materials) on extracts is generally ignored [13]. To improve process efficiency and guarantee final product quality, reliable process analytical technology (PAT) should be emphasized in the manufacturing process of traditional Chinese medicine (TCM) and dietary supplements, especially in the extraction unit. Near-infrared (NIR) spectroscopy, associated with chemometrics, is becoming a powerful tool with lots of advantages such as fast acquisition, non-invasive, non-destructive, minimization of sample preparation, and capability to handle and control a great number of industrial/technological variables which must be optimized in manufacturing processes [14–16]. NIR spectroscopy has also shown great power and gained wide acceptance in TCM [17]. The most well-known uses of NIR spectroscopy in the manufacturing process of TCM include qualitative discrimination and quantitative determination of herb materials, intermediates and products [18–21], separation monitoring [22], end point assessment of extraction process [13,23], and alcohol precipitation monitoring [24]. Obviously, chemical information in herbal materials and herbal products can be reflected in NIR spectra, and process monitoring using NIR spectroscopy technology can play an important role in process understanding. The objective of this study is to investigate the feasibility and application of NIRS in determining puerarin, daidzin, daidzein and total isoflavonoid (TIF) concentrations during the extraction process of kudzu. NIR spectroscopy calibration models were established using partial least squares (PLS) based on the NIR spectral and reference data collected from the kudzu extraction process. Three variable selection methods including correlation coefficient method, interval partial least squares (iPLS) and successive projections algorithm (SPA) were employed for variable selection and model optimization. The performance of the models were evaluated using the correlation coefficients and prediction errors in both calibration and validation steps. To our best knowledge, this research was the first to report the NIR spectroscopy method for the determination of isoflavonoids during kudzu extraction process, and could also be used as a reference to achieve similar purposes of other herb materials.
Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). Ethanol (70%, v/v) used as the extraction solvent was medicinal grade. HPLC grade acetonitrile was obtained from Fisher Scientific (Pittsburgh, PA). Water was purified by Millipore water purification device (Millipore, USA). All the other reagents were of analytical grade. Sample preparation Kudzu (8 kg) was extracted with 8-fold 70% ethanol at 85 ± 2 °C and repeated twice. The first extraction process lasted 90 min and the second lasted 60 min. Samples collected at different times were centrifuged at 8000 rpm for 10 min. The supernatant was analyzed by NIR, HPLC and UV, respectively. In order to obtain similar prediction accuracy, it was necessary to ensure uniform distribution of samples with high or low concentrations [13]. In the first extraction process, extract samples were collected after 20 min and at 5 min intervals over the subsequent 70 min. In the second extraction process, extract samples were collected after 18 min and at 7 min intervals over the subsequent 42 min. In this study, four extraction batches were carried out and 88 samples were collected. Spectra acquisition The NIR spectra were performed by an Antaris II Fourier transform near-infrared spectrophotometer (Thermo Fisher Scientific, USA). Transmission spectra of samples were collected from 10,000 to 4000 cm 1 at every 4 cm 1 path intervals. Each spectrum was obtained by averaging 32 scans and the gain was auto-optimized to increase the signal to noise ratio. To avoid error from the outer environment, all samples were equilibrated to room temperature (25 °C) prior to NIR analysis. And the humidity was also kept at ambient level in the laboratory. Determination of reference values Determination of reference values was performed immediately after the NIR measurements. After appropriate dilution with 70% ethanol, the TIF was determined using UV spectroscopy reported by Qingxin Kong [25], and the validated reversed-phase HPLC assay [26] with a slight modification was applied to the quantitative determination of puerarin, daidzin and daidzein. Chromatographic analysis was performed on an Agilent 1260 HPLC system (Agilent Corp., Santa Clara, CA, USA). A Phenomenex C18 column (250 mm 4.6 mm i.d. 5 lm) was employed. The mobile phase was composed of acetonitrile (A) and 0.05% phosphoric acid in water (B). The gradient elution had the following profile: 8–12% (A) in 0–21 min, 12–17% (A) in 21–31 min, 17–38% (A) in 31– 55 min, 38–90% (A) in 55–65 min and 90% (A) in 65–75 min, at a flow rate of 1.0 mL/min. The wavelength of UV detector was set at 278 nm and the injection volume was 10 lL. Data processing All the computations, including data collecting and loading, division of the calibration and validation set, spectral pre-processing, variable selection, chemometric regression model construction and validation were performed using MATLABÒ version 7.10 environment (Math-Works, Natick, USA) with the PLS-toolbox version 4.0 (Eigenvector Research, Inc., Wenatchee, WA) and TQ analyst software package (Version 8.0, Thermo Scientific, Madison, WI, USA).
Materials and methods Results and discussion Plant materials and reagents Reference values analysis Kudzu materials were supplied by a TCM pharmaceutical factory (Shandong Wohua Pharmaceututial Co., Ltd., Weifang, China). Puerarin, daidzin and daidzein were purchased from the National
Using the HPLC and UV spectrophotometric methods described in Section ‘Determination of reference values’, all 88 samples were
1405
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
set while the remaining samples were split into two groups, calibration and validation sets, by K–S algorithm. In this research, 60 samples were placed into the calibration set, and the remaining 28 samples were placed into the validation set. The statistical values are listed in Table 2. After division by K–S algorithm, the content values of isoflavonoids in the calibration and the validation set can cover a wide enough range, which is helpful to develop robust models.
Fig. 1. Characteristic chromatograms of the mixed standards (A) and kudzu solution sample (B). (1. Puerarin, 2. Daidzin, 3. Daidzein.)
analyzed. Representative HPLC chromatograms of the standard solution (three isoflavonoid components) and the extract solution of kudzu are shown in Fig. 1. Baseline separation for all three analytes can be achieved. Before the sample testing, both the HPLC and UV spectrophotometric methods were validated. The main methodology parameters and calibration curves of the reference HPLC and UV method are listed in Table 1. Analysis of raw NIR spectra Fig. 2 shows the NIR spectra of 88 kudzu extract samples. There are no apparent differences in the raw NIR spectra of samples due to high-degree band overlapping. However, the usable spectral regions are necessary to be identified prior to developing the calibration models. The NIR spectra exhibit intense absorption bands around 7000 cm 1 from the first O–H overtone and around 5160 cm 1 from the combination of stretching and deformation of the O–H group in water [27]. These two bands, which are typical characteristics of the NIR spectrum of an aqueous alcoholic solution [28], hold a significant influence on determination of the active constituents with near infrared spectroscopic technique because the intense bands could mask any other bands present in these spectral ranges [28,29]. The 7500–10,000 cm 1 region characterized by low intensity and low signal-to-noise ratio is assigned to the second and third overtones [27]. The 4500– 5000 cm 1 region corresponds to combinations of fingerprint absorptions with the C–H, O–H and N–H stretching modes [30]. The 5450–6100 cm 1 region, which is related to the first overtones of the C–H stretching mode [27], is useful for quantitative NIR models, although this spectral region is always immersed in the intense water band [28]. Quantitative calibration models Division of calibration and validation set The samples with maximum and minimum concentrations of puerarin, daidzin, daidzein and TIF were included in the calibration
Optimization of calibration model In this study, for the puerarin, daidzin, daidzein and TIF models, the ‘‘fingerprint region’’ (4500–5000 cm 1), and the ‘‘water regions’’ (5000–5450 cm 1 and 6600–7500 cm 1) were excluded. The spectral region actually employed was selected from 5450 to 6600 cm 1 for daidzin and daidzein; and for puerarin and TIF which have relatively high concentrations (the concentration of puerarin higher than 0.4 mg/ml and TIF higher than 0.92 mg/ml), the region of 7500–10,000 cm 1 was also employed. To confirm the correctness of the band selection, three different variable selection methods including correlation coefficient method, iPLS and SPA were performed. Using the TIF as an example, as shown in Fig. 3, the variables selected by forward iPLS are mostly distributed in the selected region. Similar cases can also be observed for the other three quality indices (QIs). In addition, different spectral pretreatments were investigated in an effort to optimize calibration performance. As for the transmitted or transflective spectra, the most frequently used pretreated method is derivative (D) calculation, including 1st D, 2nd D, or 3rd D. To avoid noise magnification caused by derivative calculation, spectra are first smoothed using the Savitzky–Golay filter (S–G) algorithm, which is a moving window method [21]. In the present study, different spectral pretreated methods were compared and selected for each calibration model. The decision on which pre-processed method to adopt was based on the cross-validation results. The selected wave-number intervals and optimized spectral pretreatment methods for the calibration models are listed in Table 3. Determination of optimal latent variable numbers The PLSR reduces the dimensionality of spectral data through the calculation of the latent variables (LVs), which explain the maximum amount of variability in the data. Every PLS calibration model has an optimal number of LVs, which are considered as the best fit for the original spectral variability. LVs lower or greater than the optimal one introduced in the model bring about the problem of ‘‘under-fitting’’ or ‘‘over-fitting’’, and both of which will decrease predictability. Cross-validation via the venetian blinds algorithm (root-mean-squared error of cross-validation, RMSECV) was used for seeking the optimal LVs. The most suitable LVs for the models of four QIs are also shown in Table 3. Calibration and validation of the quantitative models The performance of models was assessed by coefficient of calibration (Rc), coefficient of validation (Rp), root mean square error of calibration (RMSEC), root mean square errors of cross-validation (RMSECV) and root meat square errors of prediction (RMSEP). A
Table 1 Method parameters and linear calibration functions of the HPLC and UV spectrophotometric method.
a b
Compound
tR (min)
Linearity range (lg/mL)
Calibration curve
Puerarina Daidzina Daidzeina TIFb
17.10 23.06 35.41 –
2.50–500.0 0.60–120.0 0.50–40.0 0.16–11.2
Y = 3.0 10 Y = 3.0 10 Y = 2.0 10 Y = 4.3 10
Determinated by HPLC. Determinated by UV spectrophotometric.
4 4 4 2
X + 2.7 10 X + 1.3 10 X + 6.0 10 X + 1.4 10
3 3 4 3
R
Repeatability (RSD%, n = 6)
Recovery (%, n = 6)
1.0000 1.0000 1.0000 0.9999
1.15 1.27 0.93 1.33
101.2 98.6 99.7 100.5
1406
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
Fig. 2. Raw spectra of kudzu solution in the water–ethanol extraction process.
Table 2 Statistical values of the isoflavonoids content in calibration and validation sets. Compound
Puerarin Daidzin Daidzein TIF
Total sets (mg/ml)
Calibration sets (mg/ml)
Validation sets (mg/ml)
Average values
Variation ranges
Average values
Variation ranges
Average values
Variation ranges
1.98 0.31 0.57 4.64
0.40–3.42 0.05–0.63 0.09–0.94 0.92–8.58
1.99 0.29 0.55 4.60
0.40–3.42 0.05–0.63 0.09–0.94 0.92–8.58
1.95 0.34 0.61 4.74
0.58–3.00 0.12–0.53 0.18–1.02 1.60–7.40
Fig. 3. The variable regions selected by forward iPLS.
Table 3 The most suitable conditions, and performance parameters of the calibration models established via PLSR. Spectral pretreatment method
Optimum no. of LVs
R2
RMSEC (mg/mL)
RMSECV (mg/mL)
RMSEP (mg/mL)
8649–8458, 8070–7504, 6527–5951
Autoscale,1st D, SG (15, 2)
6
0.996
0.045
0.058
0.063
Daidzin
6099–5499
Autoscale,1st D, SG (15, 2)
6
0.989
0.016
0.019
0.016
Daidzein
6099–5399
Autoscale,1st D, SG (15, 2)
5
0.985
0.032
0.038
0.034
TIF
9420–9229, 6144–5951, 5563–5372
Autoscale,1st D, SG (15, 2)
7
0.997
0.095
0.127
0.145
Compound
Wavenumber interval (cm
Puerarin
1
)
good NIR calibration model should have low RMSEC and RMSEP values, high correlation coefficient (R), but small differences between RMSECV and RMSEP. According to the procedures mentioned above, the calibration models of the four QIs were established. As shown in Table 3, the
RMSEC and R for calibration of TIF are 0.095 and 0.997, respectively, and the RMSEP and RMSECV for the validation are 0.145 and 0.127, respectively. The performance of parameters of the other three QIs models are listed in Table 3. The results show that the established models give satisfactory fitting results and predic-
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
1407
Fig. 4. The correlation charts of the four calibration models ((A) TIFs; (B) Puerarin; (C) Daidzin; (D) Daidzein).
tive ability, and can be applied to monitoring the concentration of the four QIs in the water–ethanol extraction process of kudzu. Fig. 4 presents the relevance plots between reference values (xaxis) and the PLS prediction values (y-axis) for all the four QIs. Prediction accuracy is satisfactory.
Conclusion In this study, a NIR spectroscopy method is explored and established for the simultaneous determination of puerarin, daidzin, daidzein and TIF during the water–ethanol extraction process of kudzu. By means of PLSR, robust NIR models were built between transmissive spectra and reference values with high correlations and low error of predictions. Compared with the HPLC and UV methods, the presented method based on NIRS can significantly save manpower and time. The established calibration models are applying to monitor the extraction process of kudzu by our company partner, and the successful implementation can also help other factories to achieve similar purposes. Acknowledgements We are grateful for the financial support of the Major Projects of Independent Innovation Achievements of Shandong Province (No. 2010ZDZX1A0406) and Shandong Province Natural Science Foundation (No. ZR2011HM080). The authors would like to acknowledge the collaborators from Wohua Pharmaceutical Co., Ltd., for supporting this work and for permission to publish the above results. References [1] K.H. Wong, G.Q. Li, K.M. Li, V. Razmovski-Naumovski, K. Chan, Kudzu root: traditional uses and potential medicinal benefits in diabetes and cardiovascular diseases, J. Ethnopharmacol. 134 (2011) 584–607. [2] B. Yan, D.-Y. Wang, D.-M. Xing, Y. Ding, R.-F. Wang, F. Lei, L.-J. Du, The antidepressant effect of ethanol extract of radix puerariae in mice exposed to cerebral ischemia reperfusion, Pharmacol. Biochem. Behav. 78 (2004) 319– 325. [3] H. Kim, Neuroprotective herbs for stroke therapy in traditional eastern medicine, Neurol. Res. 27 (2005) 287–301. [4] Y. Wenli, Z. Yaping, S. Bo, The radical scavenging activities of radix puerariae isoflavonoids: a chemiluminescence study, Food Chem. 86 (2004) 525–529. [5] W.M. Keung, B.L. Vallee, Kudzu root: an ancient Chinese source of modern antidipsotropic agents, Phytochemistry 47 (1998) 499–506. [6] H. Yao, B. Wu, Y. Cheng, H. Qu, High throughput chemiluminescence platform for evaluating antioxidative activity of total flavonoid glycosides from plant extracts, Food Chem. 115 (2009) 380–386.
[7] Z. Zhang, T.N. Lam, Z. Zuo, Radix puerariae: an overview of its chemistry, pharmacology, pharmacokinetics, and clinical use, J. Clin. Pharmacol. 53 (2013) 787–811. [8] L. Coward, N.C. Barnes, K.D.R. Setchell, S. Barnes, Genistein, daidzein, and their beta-glycoside conjugates: antitumor isoflavones in soybean foods from American and Asian diets, J. Agric. Food Chem. 41 (1993) 1961–1967. [9] D.K.Y. Yeung, S.W.S. Leung, Y.C. Xu, P.M. Vanhoutte, R.Y.K. Man, Puerarin, an isoflavonoid derived from radix puerariae, potentiates endotheliumindependent relaxation via the cyclic AMP pathway in porcine coronary artery, Eur. J. Pharmacol. 552 (2006) 105–111. [10] P. Van Hung, N. Morita, Chemical compositions, fine structure and physicochemical properties of kudzu (Pueraria lobata) starches from different regions, Food Chem. 105 (2007) 749–755. [11] M.-H. Lee, C.-C. Lin, Comparison of techniques for extraction of isoflavones from the root of radix puerariae: ultrasonic and pressurized solvent extractions, Food Chem. 105 (2007) 223–228. [12] J.P. Fan, J. Cao, X.H. Zhang, J.Z. Huang, T. Kong, S. Tong, Z.Y. Tian, Y.L. Xie, R. Xu, J.H. Zhu, Optimization of ionic liquid based ultrasonic assisted extraction of puerarin from radix puerariae lobatae by response surface methodology, Food Chem. 135 (2012) 2299–2306. [13] Y. Wu, Y. Jin, H. Ding, L. Luan, Y. Chen, X. Liu, In-line monitoring of extraction process of scutellarein from Erigeron breviscapus (vant.) Hand-Mazz based on qualitative and quantitative uses of near-infrared spectroscopy, Spectrochim. Acta Part A Mol. Biomol. Spectrosc. 79 (2011) 934–939. [14] J.J. Workman, Review of process and non-invasive near-infrared and infrared spectroscopy: 1993–1999, Appl. Spectrosc. Rev. 34 (1999) 1–89. [15] G. Reich, Near-infrared spectroscopy and imaging: basic principles and pharmaceutical applications, Adv. Drug Deliv. Rev. 57 (2005) 1109–1143. [16] Y. Roggo, P. Chalus, L. Maurer, C. Lema-Martinez, A. Edmond, N. Jent, A review of near infrared spectroscopy and chemometrics in pharmaceutical technologies, J. Pharm. Biomed. Anal. 44 (2007) 683–700. [17] A. Rohman, A. Nugroho, E. Lukitaningsih, Sudjadi, Application of vibrational spectroscopy in combination with chemometrics techniques for authentication of herbal medicine, Appl. Spectrosc. Rev. 49 (2014) 603–613. [18] J. Lu, B. Xiang, H. Liu, S. Xiang, S. Xie, H. Deng, Application of two-dimensional near-infrared correlation spectroscopy to the discrimination of Chinese herbal medicine of different geographic regions, Spectrochim. Acta Part A, Mol. Biomol. Spectrosc. 69 (2008) 580–586. [19] Y.A. Woo, H.J. Kim, J.H. Cho, H. Chung, Discrimination of herbal medicines according to geographical origin with near infrared reflectance spectroscopy and pattern recognition techniques, J. Pharm. Biomed. Anal. 21 (1999) 407– 413. [20] Y. Feng, D. Lei, C. Hu, Rapid identification of illegal synthetic adulterants in herbal anti-diabetic medicines using near infrared spectroscopy, Spectrochim. Acta Part A, Mol. Biomol. Spectrosc. 125 (2014) 363–374. [21] W. Li, Z. Cheng, Y. Wang, H. Qu, A study on the use of near-infrared spectroscopy for the rapid quantification of major compounds in Tanreqing injection, Spectrochim. Acta Part A, Mol. Biomol. Spectrosc. 101 (2013) 1–7. [22] H. Liu, X. Zhao, T. Qi, Y.P. Qi, G.R. Fan, Establishment of the model for online monitoring of the column separation and purification process by near-infrared spectroscopy and determination of total ginsenosides in Folium Ginseng, Guang pu xue yu guang pu fen xi = Guang pu 33 (2013) 3226–3230. [23] Z. Wu, C. Sui, B. Xu, L. Ai, Q. Ma, X. Shi, Y. Qiao, Multivariate detection limits of on-line NIR model for extraction process of chlorogenic acid from Lonicera japonica, J. Pharm. Biomed. Anal. 77 (2013) 16–20. [24] Z. Wu, B. Xu, M. Du, C. Sui, X. Shi, Y. Qiao, Validation of a NIR quantification method for the determination of chlorogenic acid in Lonicera japonica solution in ethanol precipitation process, J. Pharm. Biomed. Anal. 62 (2012) 1–6. [25] K. Qingxin, Process study of extracting total isoflavones from radix puerariae by microwave-assisted extraction, J. Anhui Agric. Sci. 35 (2007) 10835.
1408
P. Wang et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 137 (2015) 1403–1408
[26] P. Wang, L. Li, H. Yang, S. Cheng, Y. Zeng, L. Nie, H. Zang, Chromatographic fingerprinting and quantitative analysis for the quality evaluation of Xinkeshu tablet, J. Pharm. Anal. 2 (2012) 422–430. [27] N. Gierlinger, M. Schwanninger, R. Wimmer, Characteristics and classification of Fourier-transform near infrared spectra of the heartwood of different larch species (Larix sp.), J. Near Infrared Spectrosc. 12 (2004) 113–119. [28] Y. Jin, Z. Wu, X. Liu, Y. Wu, Near infrared spectroscopy in combination with chemometrics as a process analytical technology (PAT) tool for on-line
quantitative monitoring of alcohol precipitation, J. Pharm. Biomed. Anal. 77 (2013) 32–39. [29] Q. Dong, H. Zang, A. Liu, G. Yang, C. Sun, L. Sui, P. Wang, L. Li, Determination of molecular weight of hyaluronic acid by near-infrared spectroscopy, J. Pharm. Biomed. Anal. 53 (2010) 274–278. [30] K.-Z. Liu, M. Shi, A. Man, T.C. Dembinski, R.A. Shaw, Quantitative determination of serum LDL cholesterol by near-infrared spectroscopy, Vib. Spectrosc. 38 (2005) 203–208.
