Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 120 (2014) 499–504

Contents lists available at ScienceDirect

Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa

Fast determination of two atractylenolides in Rhizoma Atractylodis Macrocephalae by Fourier transform near-infrared spectroscopy with partial least squares Qing-Song Shao a, Ai-lian Zhang a, Wen-Wen Ye a, Hai-Peng Guo b, Run-Huai Hu a,⇑ a b

The Research Center for the State Key Laboratory of Subtropical Silviculture, Zhejiang A & F University, Hangzhou 311300, China Zhejiang University, Hangzhou 310058, 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

 The RMSECV of the PLS models for

NIRS standard model for atractylenolide I and III.

atractylenolides I and III was 0.0387 and 0.0358.  The determination coefficient of quantitative models was 96.63 and 96.16.  The established predictions model were accurate and reliable.

a r t i c l e

i n f o

Article history: Received 6 July 2013 Received in revised form 7 October 2013 Accepted 9 October 2013 Available online 20 October 2013 Keywords: Rhizoma Atractylodis Macrocephalae Atractylenolide I Atractylenolide III FT-NIR Partial least squares method

a b s t r a c t Rhizoma Atractylodis Macrocephalae (RAM) is a commonly used food and traditional Chinese medicine (TCM), which traditionally strengthens the spleen, benefits vital energy, eliminates dampness, and promotes hidroschesis. Its primary effective constituents are polysaccharides and volatile oil, whose main components are atractylenolide I and III. Fourier transform near-infrared spectroscopy (FT-NIR) is widely used in TCM research. However, determination of atractylenolides in RAM using FT-NIR has not been described. In this study, a new method for the determination of atractylenolides I and III in RAM by NIR was established. The spectral characteristics of atractylenolides I and III were obtained by second derivative multiple scattering correction, and its chart to the original absorbance spectra. Additionally, in combination with the partial least squares (PLS) algorithm, the calibration process was performed for the quantitation of the samples. The root mean square error of cross-validation of the PLS models for atractylenolides I and III was 0.0387 and 0.0358, and the determination coefficient of quantitative models was 96.63 and 96.16, respectively. This study demonstrated that NIR spectroscopy can be used to analyze quickly and efficiently the contents of atractylenolides I and III in RAM. Ó 2013 Elsevier B.V. All rights reserved.

Introduction Rhizoma Atractylodis Macrocephalae (RAM) is the dried root of Atractylodes macrocephala Koidz., produced mainly in Zhejiang, ⇑ Corresponding author. Tel./fax: +86 571 63740809. E-mail address: [email protected] (R.-H. Hu). 1386-1425/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.saa.2013.10.035

Hebei, Hubei, Anhui, Hunan, and Jiangxi [1]. RAM cultivated in Zhejiang is one of the most famous Zhejiang Bawei medicines [2]. RAM is used as a food and medicinal herb in China and has many effects, such as strengthening the spleen, benefiting vital energy, eliminating dampness, hidroschesis, and soothing fetuses [3]. RAM contains polysaccharides and volatile oil, the latter of which makes up about 1.4% of the total compounds. The main

500

Q.-S. Shao et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 120 (2014) 499–504

Fig. 1. NIR spectra of A. macrocephala calibration sets.

components of the volatile oil are atractylol, atractylon, atractylenolide I, atractylenolide II, atractylenolide III, and biatractylolide [4]. RAM is planted and distributed extensively but its complex origin, nonstandard cultivation, and processing lead to variable quality, especially in the volatile oil [5]. There are several methods reported in the literature for the quality control of RAM, such as thin layer chromatography [6], gas chromatography–mass spectrometry [7], high-performance liquid chromatography [8], and Fourier transform infrared spectra

[9]. However, most of these methods have some disadvantages, such as tedious and complex processing for samples, long analysis time, destruction of samples, and the use of chemical reagents [10]. A reliable technique is necessary to quickly and nondestructively detect the quality of RAM. Fourier transform near-infrared spectroscopy (FT-NIR) is a fast and non-destructive technique [11] and has been widely used in quality control and evaluation of traditional Chinese medicines [12–25]. FT-NIR can give the response to transition of corresponding chemical constituents, such as O–H, N–H, and C–H [26]. FT-NIR has the advantage of not having special sample preparation or the use of chemical reagents, uses little sample, and is pollution-free, rapid, and easy to use [27]. Combined with chemometrics, FT-NIR can rapidly analyze complex samples. In this study, the contents of atractylenolides I and III were determined by near-infrared diffuse reflectance spectra combined with the partial least square (PLS), establishing an analytical method for RAM. Root mean square error of cross-validation (RMSECV) and the determination coefficient (R2) were used for atractylenolide I and III contents. Materials and methods Samples and reagents One hundred ten batches of RAM were collected for this study in 2007 from Fujian and Zhejiang Provinces in China. Before the

Table 1 Results of different pretreatments on atractylenolide I and III. Pretreatments method

Coefficient of determination R2

RMSECV (%)

Dimensions

Atractylenolide I

No pretreatment First derivative Second derivatives First derivative + vector normalization First derivative + subtract a straight line First derivative + MSC Vector normalization Maximum–minimum normalization Multiplicative scatter correction Subtract a straight line Constant offset elimination Internal standard

73.93 95.53 85.80 96.63 92.84 91.22 86.19 80.82 74.86 89.64 69.94 77.00

0.1080 0.0446 0.0796 0.0387 0.0565 0.0626 0.0784 0.0925 0.1060 0.0679 0.1160 0.1010

10 10 9 10 9 7 10 10 9 10 9 6

Atractylenolide III

No pretreatment First derivative Second derivatives First derivative + vector normalization First derivative + minus a straight line First derivative + MSC Vector normalization Maximum–minimum normalization Multiplicative scatter correction Minus a straight line Eliminate the constant offset Internal standard

92.73 96.16 90.57 91.40 94.94 91.35 92.58 90.59 93.32 92.76 93.46 88.90

0.0492 0.0358 0.0560 0.0535 0.0410 0.0537 0.0497 0.0560 0.0472 0.0491 0.0467 0.0608

10 8 6 6 7 6 10 10 10 9 10 10

Table 2 Results of different wavelength on evaluation of atractylenolide I and III. Band Atractylenolide I

Atractylenolide III

1

10,530–9689.2 cm 12,493.3–3598.7 cm 9029.6–3718.3 cm 1 10,414.3–9538.7 cm

1

1

1

6024.9–5114.6 cm 12,493.3–3598.7 cm 9029.6–3718.3 cm 1 6125.2–5454 cm 1

1

Coefficient of determination R2

Cross-validation root mean square (%)

Recommended dimensions

71.96 48.15 57.34 96.63

0.1120 0.1520 0.1380 0.0387

9 9 8 10

87.06 27.83 59.21 96.16

0.0656 0.1550 0.1243 0.0358

10 7 8 8

Q.-S. Shao et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 120 (2014) 499–504 Table 3 Results of different smoothing points on evaluation of atractylenolide I and III. Even points

Coefficient of determination R2

Crossvalidation root mean square (%)

Recommended dimensions

Atractylenolide I

5 9 13 17 21 25

87.97 92.87 95.56 96.62 96.63 96.09

0.0732 0.0564 0.0445 0.0388 0.0387 0.0417

9 9 10 10 10 10

Atractylenolide III

5 9 13 17 21 25

93.74 95.55 96.16 95.76 94.55 91.41

0.0456 0.0385 0.0358 0.0376 0.0426 0.0535

8 8 8 8 9 10

501

Spectra collection and software Samples were scanned in diffuse reflectance mode (13,000– 3500 cm 1) using a multi-purpose analyzer spectrometer (Bruker Optics, Ettlingen, Germany) with a RT-PbS detector and an internal gold background as the reference. NIR spectral data were collected using OPUS 5.5 (Bruker Optics, Ettlingen, Germany) and stored in absorbance format. After being stirred, each sample was scanned in a quartz cuvette three times at room temperature. Then, for each sample, three reflectance spectra were obtained and the average was used for the sample in calibration or prediction. The spectrum scanning range was from 12,500 cm 1 to 3600 cm 1. Both sample and background spectra were collected in absorbance. The scanning parameters were as follows: resolution of 2 cm 1, scanner velocity of 10 kHz, 32 scans for each spectrum. Spectral acquisition was conducted at ambient temperature of 25 °C and relative humidity of 30%. Scanning was performed 64 times, the resolution was 8 cm 1, and air was used as a blank control. To avoid errors from uneven samples, the sample pool was rotated 120° to record another spectrum after each recording, and the average of the four spectra was considered as the final spectrum. HPLC measurement of atractylenolide I and III All samples were analyzed by a Waters Acquity UPLC™ system (Waters, Milford, MA, USA) equipped with a binary solvent

Fig. 2. NIRS standard model for atractylenolide I and III.

spectra were collected, samples were dried and crushed by an oscillation ball mill (MM 301, Retsch, Haan, Germany). Then, the powders were sieved with a 90 lm mesh, and saved for further experiment. Forty-eight samples were selected randomly as the calibration set for modeling. The other forty-eight samples were used as the prediction set for testing the prediction accuracy the model.

Fig. 3. Relationship between R2 and dimension.

502

Q.-S. Shao et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 120 (2014) 499–504

The stock solutions of atractylenolide I and III were prepared with 0.005, 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1.0 ml, and added up to 1 ml in a volumetric flask with methanol. All stock solutions were then shaken well. Then, linear-regression analysis was conducted. The eight-point calibration curves were made at concentrations ranging from 0.71–142 lg ml 1 for atractylenolide I and 0.765– 153 lg ml 1 for atractylenolide III. The peak area and quality concentrations showed a linear relationship in the respective concentration ranges of atractylenolide I and III. The correlation coefficients of the calibration curves of atractylenolide I and III were 0.9999 and 0.9997, respectively. Spectral data pre-treatment Mathematical pre-treatments were performed to enhance the contribution of the chemical composition. By smoothing the magnitude of absorption peaks, the position of asymmetric absorption bands was shifted. After an automatic smoothing of the spectrum, the derivatives were obtained. In this experiment, the pre-treatment methods of spectra were compared and studied with OPUS5.5 software, including no pretreatment, first derivatives, second derivatives, first derivatives + vector normalization, multiplicative scatter correction (MSC), first derivatives + straight subtraction, first derivative + MSC, vector normalization, maximum-minimum normalization, straight line subtraction, constant offset elimination, and internal standardization.

Results and discussion Primary election of the spectra regions

Fig. 4. Relationship between RMSECV and dimension.

delivery pump, an autosampler, and a photo diode array detector. The system used Waters MassLynxTM software (version 4.1, Waters, Milford, MA, USA). UPLC grade acetonitrile was purchased from Merck (Darmstadt, Germany). Deionized water was prepared using a Simplicity 185 system (Millipore, Bedford, MA, USA), and analytical grade solvents for extraction and chromatography were purchased from Beijing Beihua Fine Chemicals Co., Ltd., (Beijing, China). The references, atractylenolide I and atractylenolide III, were purchased from Shanghai Shunbo Biotech Co., Ltd. The chromatographic separation was performed using a Waters Acquity UPLCTM BEH C18 column (1.0 mm  50 mm, 1.7 lm), operated at 30 °C and detected at 220 nm for atractylenolide I and III. The mobile phase consisted of acetonitrile/0.1% methanoic acid (60:40, v/v) at a flow-rate of 0.1 ml min 1. Samples were quantified by the external standard method from peak areas. Results of UPLC were used for reference in NIR analysis. The powder (0.2 g) was accurately weighed, ultrasonically extracted in 10 ml of methanol in dark brown calibrated flasks for 30 min, and then transferred into additional methanol, which was added to make up for evaporation. These solutions were stored in dark glass bottles at 4 °C. The standard solutions and the extractions were filtered through a 0.45 lm Nylon membrane microfilter (Alltech, Breda, Netherlands) and analyzed (1.0 ll) by UPLC.

Under the same spectrum acquisition conditions to ensure that the tested samples were uniform and the loading thickness are consistent, each sample was passed through the 6th pharmacopoeia screen rather than 8th pharmacopoeia screen in this experiment. There are 74 corrections near the infrared mean spectra of RAM powders in Fig. 1, which shows that the near infrared spectra of different samples are very close. Therefore, it is impossible to directly distinguish the correlation between absorbance and the content of atractylenolide I or III. In Fig. 1, the curve is smooth and no obvious changes were found with absorbances in the band range of 12,500–9000 cm 1. The end absorption was strong in the band range of 4000–3600 cm 1. Absorbance information was very abundant in the band range of 9000–4000 cm 1. Therefore, to save analysis time, the suitable analysis bands of atractylenolide I and III were selected with the optimization program of OPUS 5.5 in the band range of 9000–4000 cm 1. The model was also adjusted and optimized for these bands. Selection of the pretreatment methods In this experiment, the pretreatment methods of spectra were compared by OPUS 5.5 software. In the wavelength range of 10,414.3–9538.7 cm 1, Table 1 shows the results of different pretreatments on atractylenolide I when the smoothing points of the first and second derivative were 21. Under the same band and smoothing points, the best prediction of atractylenolide I content was found using the first and second derivative pretreatment method, which gets the maximum R2. Similarly, in the wavelength range of 6125.2–5454 cm 1, Table 1 shows the results of different pretreatments on atractylenolide III when the smoothing points of the first and second derivative were 13. Under the same band and smoothing points, the best prediction of atractylenolide III content

503

Q.-S. Shao et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 120 (2014) 499–504 Table 4 Result of prediction of atractylenolide I. Sample numbers

UPLC measurements (lg mg 1)

NIR measurements (lg mg 1)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

0.784 0.852 0.893 0.834 0.811 0.969 0.737 0.748 0.667 0.607 0.525 0.545 0.515 0.476 0.565 0.694 0.661 0.348 0.275 0.287 0.279 0.344 0.241 0.614 0.406 0.440 0.472 0.630 0.705 0.772 0.267 0.256

0.7757 0.7941 0.8903 0.8300 0.7842 0.9215 0.7342 0.7721 0.7048 0.6592 0.5869 0.5513 0.5184 0.4359 0.5531 0.6688 0.6775 0.3797 0.2493 0.2779 0.3005 0.3508 0.2581 0.5749 0.4330 0.4270 0.4685 0.6178 0.6975 0.8052 0.2774 0.2555

Deviation 0.0083 0.0579 0.0027 0.0040 0.0268 0.0475 0.0028 0.0241 0.0378 0.0522 0.0619 0.0063 0.0034 0.0401 0.0119 0.0252 0.0165 0.0317 0.0257 0.0091 0.0215 0.0068 0.0171 0.0391 0.0270 0.0130 0.0035 0.0122 0.0075 0.0332 0.0104 0.0005

Markov distance

Residual

F value

F Prob

0.050 0.081 0.167 0.123 0.066 0.183 0.058 0.253 0.043 0.060 0.065 0.095 0.101 0.460 0.061 0.197 0.127 0.137 0.170 0.104 0.213 0.193 0.203 0.187 0.267 0.180 0.207 0.230 0.233 0.463 0.307 0.337

0.0024 0.0023 0.0025 0.0034 0.0027 0.0033 0.0029 0.0061 0.0027 0.0027 0.0031 0.0031 0.0037 0.0028 0.0034 0.0064 0.0056 0.0038 0.0039 0.0040 0.0040 0.0043 0.0067 0.0031 0.0041 0.0039 0.0048 0.0067 0.0052 0.0085 0.0041 0.0039

0.3107 0.2675 0.3397 0.6130 0.3757 0.5790 0.4610 0.8393 0.3947 0.3775 0.4957 0.5203 0.7450 0.4217 0.5980 0.5475 0.5733 0.7660 0.8095 0.8477 0.8583 0.9650 0.6665 0.4960 0.8823 0.8175 0.5716 0.6865 0.9163 0.9830 0.8887 0.8063

0.421 0.393 0.436 0.563 0.456 0.545 0.496 0.640 0.467 0.460 0.516 0.526 0.600 0.481 0.558 0.586 0.440 0.615 0.629 0.638 0.643 0.669 0.602 0.515 0.647 0.626 0.428 0.625 0.560 0.749 0.650 0.627

Markov distance

Residual

F value

F Prob

0.032 0.034 0.106 0.080 0.045 0.193 0.042 0.203 0.032 0.050 0.047 0.067 0.046 0.140 0.054 0.083 0.133 0.115 0.130 0.110 0.079 0.293 0.210 0.115 0.280 0.520 0.380 0.190 0.247 0.487 0.243 0.310

1.367 1.550 1.717 2.387 2.280 2.180 1.497 3.887 1.913 1.763 2.295 1.553 2.240 2.357 1.760 2.703 3.070 2.155 1.817 2.503 3.483 2.690 2.090 2.400 2.535 2.965 3.535 2.977 2.723 2.640 2.747 2.560

0.3173 0.4263 0.5007 0.7950 0.7820 0.8077 0.3813 0.4947 0.6337 0.5290 0.7350 0.4103 0.6390 0.7035 0.5267 0.8573 0.6333 0.8115 0.5703 0.550 0.3628 0.7481 0.7430 0.9080 0.5656 0.8950 0.7403 0.7596 0.7565 0.6046 0.7695 0.7990

0.424 0.472 0.518 0.669 0.649 0.629 0.460 0.583 0.565 0.529 0.606 0.476 0.574 0.658 0.530 0.732 0.791 0.619 0.541 0.632 0.551 0.727 0.609 0.676 0.413 0.653 0.629 0.534 0.468 0.720 0.740 0.705

Table 5 Result of prediction of atractylenolide III. Sample numbers

UPLC measurements (lg mg 1)

NIR measurements (lg mg 1)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

0.577 0.681 0.779 0.659 0.647 0.946 0.490 0.568 0.655 0.576 0.594 0.696 0.636 0.605 0.634 0.595 0.873 0.332 0.336 0.315 0.364 0.298 0.186 0.591 0.326 0.480 0.495 0.627 0.815 0.724 0.361 0.379

0.5697 0.7052 0.7521 0.6298 0.6755 0.9001 0.4572 0.5778 0.6307 0.6185 0.6705 0.6795 0.6034 0.6019 0.6477 0.6496 0.8333 0.3578 0.3149 0.3017 0.3984 0.3459 0.1919 0.5671 0.3204 0.5175 0.4835 0.6051 0.7898 0.723 0.3715 0.3578

Deviation 0.0073 0.0242 0.0269 0.0292 0.0285 0.0459 0.0328 0.0098 0.0243 0.0425 0.0765 0.0165 0.0326 0.0031 0.0137 0.0546 0.0397 0.0258 0.0211 0.0133 0.0344 0.0479 0.0059 0.0239 0.0056 0.0375 0.0115 0.0219 0.0252 0.0010 0.0105 0.0212

504

Q.-S. Shao et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 120 (2014) 499–504

was found using the first derivative pretreatment method, which gets the maximum R2. Selection of the analysis band To better determine the analysis band in this experiment, the coefficient of determination (R2) and the cross-examination RMSECV of different wave bands with the same pretreatment method was compared. Table 2 shows the results of different pretreatments on atractylenolide I when NSP of the first derivative + vector normalization were 21. Table 2 also shows the results of different pretreatments on atractylenolide III when the NSP of the first derivative was 13. From these results, the optimization parameters recommended by the instrument, such as R2 and RMSECV, are not satisfactory. To get a better model, analysis was compared on the intervals near the recommended band. Based on this analysis, the best spectra area of optimization model was finally determined. Selection of the smoothing points Under the same pretreatment and band conditions, different even points will still lead to differences of coefficient of determination (R2), cross-validation root mean square (RMSECV), and recommended dimensions. To optimize the best even points, the atractylenolide I was evaluated in the first derivative + vector normalized spectral pretreatment methods when the NIR spectra were scanned in the range of 10,414.3–9538.7 cm 1. When the IR spectra were scanned in the range of 6125.2–5454 cm 1, the atractylenolide III was evaluated in the first derivative pretreatment methods. Results of different NSP on evaluation atractylenolide I and III are listed in Table 3. Evaluation results According to comparative study, the best standard model for atractylenolide I was in a NIR spectra range of 10,414.3– 9538.7 cm 1, 21 even points and 10 dimensions. For atractylenolide III, the best standard model was in a NIR spectra range of 6125.2–5454 cm 1, 13 even points and eight dimensions. With these optimized parameters, a NIRS analysis model for atractylenolides I and III was established. The prediction and the true value, RMSECV and dimension, and R2 and dimension of atractylenolide I and III are listed in Figs. 2–4, respectively. The figures show that the prediction and true value of atractylenolide I and III show a good correlation, in which points are evenly distributed on both sides of the line. There is no complete positive- or negative-correlation between dimension and R2 or RMSECV. With increasing dimension, RMSECV will gradually become smaller. Therefore, if the dimension is too small, a model of the quality evaluation can be degraded. Otherwise, this could cause over-fitting of the model and may not reflect the true information of the samples. In conclusion, the data for R2 and dimension and RMSECV and dimension provide us with the best way to choose dimension. Using the two models optimized to predict 32 batches of atractylenolide medicinal materials, three samples were randomly taken from each batch of RAM. The forecast average of the three samples in each batch was taken as the final prediction of per batch, and the results are shown in Tables 4 and 5. The content range of atractylenolide I is 0.241–0.969 lg mg 1, the absolute value of the deviation between true value and the prediction is 0.0005–0.0619 lg mg 1. The spectrum residual value ranges from 0.0023 to 0.0085. For atractylenolide III, the content range is 0.186–0.946 lg mg 1, the absolute value of the deviation between true value and the prediction is 0.0010–0.0765 lg mg 1, and the

spectrum residual value ranges from 1.367  10 5 to 3.887  10 5, which is smaller than that of atractylenolide I. Therefore, the reproducibility of the atractylenolide III spectrum is better than that of atractylenolide I, which agrees with the RMSECV value in the model of atractylenolide I and III. Although the R2 of atractylenolide I is higher than that of atractylenolide III, the RMSECV value of atractylenolide III is larger. The RMSECV value directly affects the quality of the model, that is, the smaller the RMSECV value, the more quality the model. Therefore, the standard model of atractylenolide III has higher accuracy. Both F value and F (F Prob) are less than 0.99, which means that the two standard models established are quite reliable in their concentration ranges. Conclusion In this study, particle size of medicinal materials and loading quantity of samples were investigated. With the OPUS5.5 optimization program, different wavebands for atractylenolide I and III were chosen. The results of different pretreatment methods, band analysis, and smoothing points were compared, and optimization models of atractylenolide I and III were established. This study demonstrated that NIR spectroscopy with PLS could be used for the quality control of atractylenolides I and III in RAM quickly and efficiently. Acknowledgments We would like to thank our colleagues Guohua Xia for their help with measurement of FT-NIR spectra using their spectrometers. This study was supported financially by Science and Technology Innovation Program for Students of Zhejiang (Xinmiao Talent Scheme) (2008R40G2100088). References [1] Y.Q. Liu, Q. Cai, Chin. J. Pharm. Anal. 32 (2012) 1249–1252. [2] L.J. Tong, X.P. Wang, Strait Pharm. J. 25 (2013) 36–39. [3] Editorial Committee of China Pharmacopoeia, China Pharmacopoeia, Part I, China Chemical Industry Press, Beijing, 2010. p. 95. [4] X.J. Xu, M. Chen, L. Zong, F.J. Hou, Y.G. Jie, Chin. Hosp. Pharm. 31 (2011) 1422– 1424. [5] C. Yu, Y. Yang, S.Y. Liu, Z.C. Bai, Chin. Tradit. Herbal Drugs 7 (2003) 766–769. [6] Y. Yan, X.M. Cheng, G.X. Chou, Lishizhen Med. Mater. Med. Res. 20 (2009) 1702–1703. [7] Q.X. Cui, Y. Dong, H.S. Wang, Chin. J. Pharm. Anal. 26 (2006) 124–126. [8] K.L. Yan, X.Q. Zhu, L. Zhao, Chin. J. Pharm. Anal. 31 (2011) 758–763. [9] Q.H. Hong, C.G. Cheng, Z.F. Cheng, D.T. Li, Spectrosc. Spect. Anal. 27 (2007) 283–286. [10] N.L. Yang, Y.Y. Cheng, Y.J. Wu, Acta Chim. Sin. 61 (2003) 393–398. [11] Y.D. Liu, X.D. Sun, A.G. Ouyang, LWT-Food Sci. Technol. 43 (2010) 602–607. [12] Y. Gao, Y.F. Chai, Y.T. Wu, Chin. Tradit. Pat. Med. 21 (2005) 1440–1443. [13] Y.H. Wu, Y. Zheng, Q.Q. Li, Vib. Spectrosc. 55 (2011) 201–206. [14] B.K. Huang, B. Zhu, H.C. Zheng, L.P. Qin, Q.Y. Zhang, J. Chin. Med. Mater. 25 (2002) 874–875. [15] Y.J. Wu, W. Li, B.R. Xiang, P. Shu, N.H. Wang, C.Q. Yuan, J. Chin. Med. Mater. 24 (2001) 26–28. [16] Y.A. Woo, H.J. Kim, J.H. Cho, H. Chung, J. Pharmaceut. Biomed. 21 (1999) 407– 413. [17] Y.A. Woo, H.J. Kim, K.R. Ze, J. Pharmaceut. Biomed. 36 (2005) 955–959. [18] G.X. Ren, F. Chen, J. Agric. Food Chem. 47 (1999) 2771–2775. [19] L.L. Liu, W.X. Xing, N. Jia, Second Mil. Med. Univ. 23 (2002) 1230–1232. [20] L. Wang, Y. He, Z.C. Qiu, X.R. Wang, X.C. Li, Spectrosc. Spect. Anal. 25 (2005) 1397–1399. [21] N.L. Yang, H.B. Zhai, Y.Y. Cheng, J. Zhejiang Univ. Eng. Sci. 36 (2002) 463– 466. [22] B. Chen, L.L. Zhao, H.J. Li, Y.L. Yan, Spectrosc. Spect. Anal. 22 (2002) 976–979. [23] Y. Bai, Z.X. Yu, S.Q. Sun, F.Y. Zhu, D.Z.H. Chen, Chin. Tradit. Herbal Drugs 36 (2005) 1391–1394. [24] J.P. Fan, Z.L. Zhang, L.Y. Zhang, Y.T. Zhang, Acad. J. Second Mil. Med. Univ. 26 (2005) 1194–1195. [25] Y.N. Shao, H. Yong, S.J. Feng, Food Res. Int. 40 (2007) 835–841. [26] L. Wang, F.S.C. Lee, X.R. Wang, LWT-Food Sci. Technol. 40 (2007) 83–88. [27] M.I. Kaniu, K.H. Angeyo, A.K. Mwala, Anal. Chim. Acta 729 (2012) 21–25.

Fast determination of two atractylenolides in Rhizoma Atractylodis Macrocephalae by Fourier transform near-infrared spectroscopy with partial least squares.

Rhizoma Atractylodis Macrocephalae (RAM) is a commonly used food and traditional Chinese medicine (TCM), which traditionally strengthens the spleen, b...
1MB Sizes 0 Downloads 0 Views