close

Вход

Забыли?

вход по аккаунту

?

00032719.2017.1385618

код для вставкиСкачать
Analytical Letters
ISSN: 0003-2719 (Print) 1532-236X (Online) Journal homepage: http://www.tandfonline.com/loi/lanl20
Characterization of Paris polyphylla var.
yunnanensis by Infrared and Ultraviolet
Spectroscopies with Chemometric Data Fusion
Yuangui Yang & Yuanzhong Wang
To cite this article: Yuangui Yang & Yuanzhong Wang (2017): Characterization of Paris polyphylla
var. yunnanensis by Infrared and Ultraviolet Spectroscopies with Chemometric Data Fusion,
Analytical Letters, DOI: 10.1080/00032719.2017.1385618
To link to this article: http://dx.doi.org/10.1080/00032719.2017.1385618
Accepted author version posted online: 20
Oct 2017.
Submit your article to this journal
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=lanl20
Download by: [University of Florida]
Date: 27 October 2017, At: 21:29
Supertitle: Spectroscopy
Characterization of Paris polyphylla var. yunnanensis by
Downloaded by [University of Florida] at 21:29 27 October 2017
infrared and ultraviolet spectroscopies with chemometric
data fusion
Yuangui Yang
Institute of Medicinal Plants, Yunnan Academy of Agricultural Sciences, Kunming, China
Yuanzhong Wang
Institute of Medicinal Plants, Yunnan Academy of Agricultural Sciences, Kunming, China
Address correspondence to Yuanzhong Wang. Tele.: +8687165033575. Fax: +8687165033441.
E-mail: boletus@126.com
ABSTRACT
Paris polyphylla var. yunnanensis has been used for its anti-tumor, anthelmintic and
hemostatic properties. In this investigation, Fourier transform infrared and ultraviolet
spectroscopy combined with chemometrics were used for qualitative analysis of P. polyphylla
var. yunnanensis from different geographical origins in Yunnan Province. A total of 82 samples
1
for each region were divided into 57 in the calibration set and 25 in the validation set by
Kennard-Stone algorithm. Support vector machine and partial least square discrimination on the
basis of Fourier transform infrared, ultraviolet, low- and mid-level data fusion were investigated.
Different pretreatments were compared for the appropriate model. The results indicated that the
combination of Savitzky-Golay (11), second derivative and standard normal variation have the
Downloaded by [University of Florida] at 21:29 27 October 2017
best performance for support vector machine and partial least square discrimination with the
lowest root mean square error of estimation and root mean square error of cross validation and
the highest cross validation accuracy rate. The accuracies of calibration and validation for
mid-level data fusion in the model of support vector machine were 84.21% and 96% for the
partial least square discrimination values of 96.49% and 84%, which was better performance
than a single technique or low-level data fusion for the classification. Moreover, the chemical
information of sample collected from Kunming and Xishuangbanna was distinguishable from the
others. These results provide a rapid and robust strategy for quality control of P. polyphylla var.
yunnanensis for further analysis.
KEYWORDS: chemometrics, data fusion, Fourier transform infrared spectroscopy, Paris
polyphylla var. yunnanensis, ultraviolet spectroscopy
Received 24 August 2017; Accepted 24 September 2017.
Introduction
2
Paris polyphylla var. yunnanensis, as the material of Chinese patent Yunnan Bai Yao, has
been used for treating of hemorrhage and traumatic injury. Most P. polyphylla var. yunnanensis
grows at elevations above 1400m with pine, bamboo and grass in Yunnan, Sichuan and Guizhou
Province, Southwestern China (Li 1998). Various growing conditions lead to variable chemical
profiles. It was reported in a previous study that chemical content of P. polyphylla var.
Downloaded by [University of Florida] at 21:29 27 October 2017
yunnanensis in different region of Yunnan varied significantly, and that total saponins in the
south were higher than that in the central district (Yang, Jin, et al. 2017). However, the detailed
geographical origins for quality assessment are scarce. The comprehensive analysis was not
discussed in the investigation, either.
Most studies concentrate on determination of the steroid saponins of P. polyphylla var.
yunnanensis by high performance liquid chromatography in previous studies (Man et al. 2010;
Zhang et al. 2010; Yin et al. 2013). The disadvantage of this method include time consuming and
laborious protocols and few bioactive markers (Yang, Zhang, et al. 2017). As a vibrational
spectroscopic technique, Fourier transform infrared spectroscopy can respond to the chemical
bond (O-H, C-H, C = O, C = C and C = N) when scanning samples in the wavenumbers ranging
from 4000 to 400 cm1 (Lohumi et al. 2015). With the rapid, non-destructive characteristics, it
has been used for identification of different regions and harvesting time in P. polyphylla var.
yunnanensis (Zhao, Zhang, Yuan, Shen, Li, et al. 2014; Yang et al. 2016).
3
Zhao, Zhang, Yuan, Shen, Hou, et al. (2014) established a rapid identification method of
P. polyphylla from Guizhou, Guangxi and Yunnan Province. The results indicate that the method
can provide accuracy in identifying sample from different origin areas. At the same time,
ultraviolet absorption spectra of sample with different type compound at the measurement
wavelengths 190–300 nm could distinguish P. polyphylla var. yunnanensis from different areas
Downloaded by [University of Florida] at 21:29 27 October 2017
(Zhang et al. 2012).
Yang, Jin, et al. (2017) has shown that partial least square-discriminant analysis with
ultraviolet spectroscopy could separate P. polyphylla var. yunnanensis in three different regions
of Yunnan Province. It is known that long wavelength by Fourier transform infrared and short
wavelength by ultraviolet always response to different chemical information of sample.
Recently, Fourier transform infrared combined with ultraviolet spectroscopy have been used in
identifying of food, beverages and drugs (Biancolillo et al. 2014; Qi et al. 2017; Yao et al. 2017).
However, they were not applied to discriminate P. polyphylla var. yunnanensis collected from
different geographical origins.
Data fusion using two or more techniques always has better performance than a single
technique in the food or herb medicine identification (Dupuy et al. 2010; Márquez et al. 2016). It
always has three levels: low-, mid- and high-level when the combination of data. In the low-level
fusion, signals measured by the different instruments are concatenated into a single matrix.
Mid-level fusion, called feature level fusion, concatenates some relevant features from each data
4
for multivariate classification. High-level fusion (also called decision level fusion) combines the
results when predictions are calculated from each single technique to obtain the final decision.
In the literatgure, low- and mid-level data fusion are frequently used to evaluate food and
herbal medicine (Casian et al. 2017; Li et al. 2017). For example, it is reported that data fusion
strategy can improve the classification capacity using ultraviolet and Fourier transform infrared
Downloaded by [University of Florida] at 21:29 27 October 2017
spectroscopy for the discrimination of Gentiana rigescens (Qi et al. 2017). Moreover, Biancolillo
et al. (2014) has showed that the best fusion protocol, mid-level data fusion could improve the
classification performance in the calibration and validation for identification of Italian craft beer.
However, few investigations have been carried out by combining Fourier transform infrared and
ultraviolet spectroscopy for qualitative analysis of P. polyphylla var. yunnanensis. In addition,
those two techniques coupled to low and mid-level data fusion are absent in previous analysis.
The aim of this study is to establish a qualitative analysis method using data fusion (lowand mid-level) based on Fourier transform infrared and ultraviolet spectroscopy, associated with
support vector machine and partial least square discriminat analysis for identifying the different
geographical origin of P. polyphylla var. yunnanensis in Yunnan Province. Firstly, pretreatment
including multiplicative scatter correction, standard normal variate, second derivative and
Savitzky-Golay were compared in different strategies. Then, using a single technique, low- and
mid-level data fusion classification performance were compared by support vector machine and
partial least square discriminat analysis.
5
Experimental
Plant materials and reagents
A total of 82 wild P. polyphylla var. yunnanensis were collected from Kunming, Nujiang,
Wenshan, Xishuangbanna and Yuxi City of Yunnan Province, China. All samples were grown in
Downloaded by [University of Florida] at 21:29 27 October 2017
six years. They were collected in 2016. The detailed information on growth is listed in Table 1.
The identification of samples was conducted by Hang Jin (Institute of Medicinal Plants, Yunnan
Academy of Agricultural Sciences). The processed rhizome was passed through 60 meshes and
stored in plastic bags until further analysis.
Methanol (analytical-grade) was purchased from Kemiou Chemical Reagent (Tianjin,
China). Potassium bromide (spectroscopic grade) was provided by Fengchuan Fine Chemical
Research Institute (Tianjing, China).
Sample preparation and standard solution
For Fourier transform infrared spectroscopy, each sample (1.5 mg) and potassium
bromide (100 mg) were weighted accurately and mixed in the agate mortar. Each sample was
analyzed three times under 65% relative humidity, and the average spectra was calculated for the
further analysis. To investigate ultraviolet spectroscopy, methanol (10 mL) added to the
medicinal powder (0.1 g) in a test tube with sonication for 30 min. The suspension was passed
6
through filter paper and collected into the tube. The blank methanol solution was carried out to
deduct the baseline drift and noise.
Fourier transform infrared and ultraviolet spectra acquisition and preprocessing
Fourier transform infrared spectra were recorded with 64 scans in the range of 4000–400
Downloaded by [University of Florida] at 21:29 27 October 2017
cm1 at the resolution of 4 cm1 using a Perkin-Elmer Fourier transform infrared spectrometer
(Norwalk, CT, USA) equipped with a deuterated triglycine sulfate detector. The spectra data
were analyzed with Ominic 8.0 (Thermo Fisher Scientific, USA).
Ultraviolet spectra were collected by UV-2550 spectrophotometer (Shimadzu, Tokyo,
Japan) in the absorption ranging from 190 to 450 nm with sampling interval of 0.5 nm and slit
width of 2.0 nm. The data were handled with an ultraviolet probe (version 2.34, Shimadzu,
Japan).
Different spectral pretreatments could enhance resolution ratio of peak and eliminate
signal noise. Multiplicative scatter correction and standard normal variate may reduce particle
size effect which were related to the scattering in spectra (Dhanoa et al. 1994). Second derivative
processingwould eliminate baseline drifts and enhance small spectral features (Shen et al. 2012).
Savitzky-Golay conditioning was used to resolve the overlapping signals and enhance signal
properties, as well as to avoid enhancing the noise by second derivative (Zimmermann and
Kohler 2013).
7
Data fusion
The advantages of low-level fusion were a simple protocol and the ability to obtain the
correlations between variables. It concatenated data from all source into a single matrix that
included many rows (samples analyzed) and columns (variables measured).
Downloaded by [University of Florida] at 21:29 27 October 2017
In this study, variables of Fourier transform infrared and ultraviolet spectra based on
different pretreatments were concatenated for the final classification. Mid-level fusion could not
only reduce the data dimensionally, but filter signal noise and enable interpretation of the results.
The study has extracted some relevant features which are selected from principal component
analysis on the basis of Fourier transform infrared and ultraviolet spectra separately, and then
concatenate them into a single data model that is used for the further classification (Biancolillo et
al. 2014). Finally, low- and mid-level data fusion were combined to classify the sample of
different geographical origin in P. polyphylla var. yunnanensis.
Partial least square discriminat analysis and support vector machine
A total of 82 samples for each region were divided into 2/3 calibration set and 1/3
validation set by Kennard-Stone algorithm for partial least square discrimination analysis and
support vector machine classification (Saptoro, Tadé, and Vuthaluru 2012). Support vector
machine is a nonlinear supervised learning technique which was first developed by Vapnik
(Vapnik 2013). Two factors, kernel function and its parameters always play an important role in
support vector machine analysis. Four kernel functions: linear, polynomial, sigmoid and radial
8
basis function, were investigated, and radial basis function was more developed than others
(Zhang et al. 2008).
In the support vector machine approach, regularization parameter C and radial basis
function kernel parameter gamma, as the essential factors, were selected in the range of 220 to
220 (Sun et al. 2017). Grid search technique was applied to search for the best C and gamma in
Downloaded by [University of Florida] at 21:29 27 October 2017
this investigation and the best C and gamma with seven-fold cross validation were used to obtain
the highest accuracy rate in the validation set (Bergner et al. 2012). The partial least square
discrimination analysis was a linear supervised classification tool, which could reduce the
dimension of complex data and evaluate whether a sample belongs to a class (Li et al. 2016).
Root mean square error of estimation and root mean square error of cross validation searched for
the best performance in the model. The parameter Q2 was employed to estimate the predicted
capability of model.
Software
The preprocessing of spectra data, data fusion and support vector machine were
conducted using Matlab (version R2014a, Math Works, USA). Partial least square discrimination
analysis was obtained by Simca-P + (version 13.0, Umetrics, Sweden).
Results and Discussion
Method validation
9
The method of ultraviolet spectroscopy was validated. Precision was analyzed by the
same sample solution in six replicates with relative standard deviations less than 3.2%. The
repeatability of sample was conducted by determining six independent sample; the results
indicate that relative standard deviations of common peaks were less than 4.2%. Stability was
determined by a single sample solution at 0, 2, 4, 8, 12, 24 h. The results indicated that all
Downloaded by [University of Florida] at 21:29 27 October 2017
analytes were stable within 24 h (relative standard deviations less than 3.3%).
In the Fourier transform infrared spectroscopy measurements, precision was calculated
by five consecutive scan with a sample. The results indicate that the correlation coefficient was
0.9997–0.9999 with a relative standard deviation 0.01%. Stability was carried out by ten minutes
per sample and five consecutive scans. The correlation coefficient ranged from 0.9994–0.9996
and the relative standard deviation was 0.01%. Repeatability was evaluated by five tablets as a
sample for scanning. The results showed that correlation coefficient was between 0.9682 and
0.9858 and therelative standard deviation was 0.88%.
Ultraviolet and Fourier transform infrared spectral features and pretreatment
The raw of ultraviolet spectra ranging from 200 to 400 nm for 82 samples of different
regions are shown in Figure 1. Significant absorption differences were 200 to 350 nm, especially
from 200–300 nm which agreed with dioscin (210 nm), paritriside C and D (243, 250, 260 nm),
and paris sapinins VII (239 nm) in P. polyphylla var. yunnanensis according to a previous
investigation (Nohara et al. 1973; Wu et al. 2013). After preprocessing spectra with the second
10
derivative, standard normal variate and Savitzky-Golay (11 points) smoothing (Figure 2),
besides 200–220 nm, absorption at 266, 271, 277, 288, 295 and 301 nm had a great influence on
the of geographical origin of P. polyphylla var. yunnanensis.
Fourier transform infrared spectra collected in the range of 4000–400 cm1 are shown in
Figure 3. Some characteristic common peaks of spectra were used by establishing Fourier
Downloaded by [University of Florida] at 21:29 27 October 2017
transform infrared spectra database for P. polyphylla var. yunnanensis. A broad band at 3394
cm1 is due to the stretching vibration peak of O-H. The peak at 2932 cm1 is mainly attributed to
the absorption of methylene group. Absorption at 1741 cm1 is assigned to C = O bending or C =
C stretching of flavonoids. Peaks at 1402 and 1373 cm1 are due to plane deviational vibrations
of methylene. Absorption bands at 1153, 1050, 930 and 861 cm1 are the characteristic of
saponins skeletal vibration (Singh, Thakur, and Schulten 1980).
As seen in Figure 4, the preprocessing Fourier transform infrared spectra (second
derivative, standard normal variate and Savitzky-Golay (11 points) smoothing) show that P.
polyphylla var. yunnanensis of different regions have various chemical compositions when
scanning samples from 1800–400 cm1. Ultraviolet and Fourier transform infrared spectral
features have certified that chemical compositions in P. polyphylla var. yunnanensis of different
region were different, although the effects of geographical origin were not well illustrated. It is
necessary that ultraviolet and Fourier transform infrared spectroscopy are combined with
chemometrics for the further analysis.
11
As shown in Table 2, different pretreatments have the various performance values in the
model. For the support vector machine, the best performance of model has high best cross
validation accuracy rate. The worst performance of model by multiplicative scatter correction
and standard normal variate has the lowest cross validation accuracy rate. In the partial least
square discriminant analysis, the appropriate model always has the low root mean square error of
Downloaded by [University of Florida] at 21:29 27 October 2017
estimation and root mean square error of cross validation. It can be seen that ultraviolet and
Fourier transform infrared have similar results. Multiplicative scatter correction and standard
normal variate have the lowest performance with high values of root mean square error of
estimation and root mean square error of cross validation. The combination of Savitzky-Golay
(11), second derivative and standard normal variate have the best performance in model of
partial least square discriminant analysis and support vector machine with the lowest root mean
square error of estimation and root mean square error of cross validation, the highest cross
validation accuracy rate.
Data fusion of ultraviolet and Fourier transform infrared spectroscopy
In the support vector machine classification, parameters C and gamma were calculated by
grid search method with a seven-fold cross validation. The optimal region was screened by the
full region of grid search (220–220) initially, then the best parameters C and gamma were
selected with the step size of 20.5 and used to calculate the accuracy rate of calibration and
validation. The results are shown in Table 3.
12
For ultraviolet spectroscopy, the best parameters C and gamma were 32 and 0.0055, the
accuracy value of calibration and validation were calculated to be 77.59% and 91.67% based on
the best parameters C and gamma. Compared with ultraviolet spectroscopy, Fourier transform
infrared spectroscopy had better performance. The value of parameter C based on ultraviolet was
sixteen times higher than for Fourier transform infrared, although they have the same accuracy
Downloaded by [University of Florida] at 21:29 27 October 2017
rate of validation, and accuracy rate of calibration of ultraviolet was higher than that of Fourier
transform infrared.
It is known that the lower value of parameter C, the lower the training error and model
complexity (Yang, Liu, et al. 2017). Thereby, in the support vector machine, the results indicate
that Fourier transform infrared had better performance than ultraviolet in the classification of P.
polyphylla var. yunnanensis from different geographical origin. In low-level data fusion, it had
the high accuracy rate in the calibration and validation, while the parameter C was 1448.15 with
high training error. Nineteen and twenty-two relevant features of ultraviolet and Fourier
transform infrared based on principal component analysis were selected for the mid-level data
fusion. As shown in Figures 5 and 6, the best parameter C and gamma were 2.8284 and 0.0884,
with accuracies of 84.21% and 96% for the calibration and validation. One of twenty-five
samples in validation set had the wrong classification. The results implied that mid-level data
fusion combined with support vector machine gave much better classification performance than a
single technique for distinguishing P. polyphylla var. yunnanensis based on geographical origin.
13
In the partial least square discrimination analysis, the opposite results were obtained;
ultraviolet spectroscopy had higher accuracy rate in the calibration and validation better
performance than Fourier transform infrared spectroscopy. The low accuracy rate of calibration
model in the low-level data fusion was coincident with support vector machine for the high
training error. The best performance for classification was mid-level data fusion with highest
Downloaded by [University of Florida] at 21:29 27 October 2017
accuracy rate in the calibration and validation among all models. We found that mid-level data
fusion had the best performance whether in the support vector machine or partial least square
discriminat analysis. When compared with support vector machine and partial least square
discriminat analysis, support vector machine always had the better performance than partial least
square discriminat analysis with high accuracy rate in the calibration and validation in single
technique of ultraviolet, Fourier transform infrared and mid-level data analysis.
Discrimination of P. polyphylla Var. yunnanensis based on geographical origin
The two-dimension score plot of partial least square discrimination analysis for P.
polyphylla var. yunnanensis in different geographical origin is displayed in Figure 7. The
contribution of the first two partial least squares components was 66.63%. It covered the most of
chemical information of the samples. Moreover, the parameter Q2 (0.5281) was more than 0.5
which suggested good predictive capacity of this model (Yu et al. 2013). Samples from Kunming
and Xishuangbanna were different from the others and were distributed in the positive and
negative principal compound 1. Chemical constituents varied significantly with the dispersive
14
spot of Kunming. The reason may be sample collected from different region in various growing
conditions. Chemical information collected from Xishuangbanna different from others was
consistent with the results in the previous study that the total saponins in Xishuangbanna was
higher than from other areas of Yunnan Province (Yang, Jin, et al. 2017). In addition, samples in
Wenshan, Nujiang and Yuxi were similar in terms of chemical components. The results agreed
Downloaded by [University of Florida] at 21:29 27 October 2017
with support vector machine analysis that a sample from Wenshan was classified into Yuxi in
Figure 6. Therefore, it had bad performance for classification of those three regions in this
model. The appropriate model needs further investigation.
Conclusions
In present study, a rapid qualitative analysis method combining Fourier transform
infrared and ultraviolet spectroscopy was used for identifying P. polyphylla var. yunnanensis
from different geographical origins in Yunnan Province. Different pretreatments was screened,
and the combination of Savitzky-Golay (11), second derivative and standard normal variate had
the best performance in model of support vector machine and partial least square discriminant
analysis with the lowest root mean square error of estimation and root mean square error of cross
validation with the highest cross validation accuracy rate.
Low- and mid-level data fusion based on Fourier transform infrared and ultraviolet
spectroscopy, associated with support vector machine and partial least square discriminat
analysis, were investigated. Mid-level data fusion in the support vector machine gave accuracies
15
of 84.21% and 96% in the calibration and validation set. Partial least square discrimination
analysis provided accuracies of 96.49% and 84% in the calibration and validation set, which
were better classification performance than a single technique for distinguishing P. polyphylla
var. yunnanensis of different geographical origins. Support vector machine in the calibration and
validation set with accuracies 71.93% and 92% for Fourier transform infrared, 77.59% and
Downloaded by [University of Florida] at 21:29 27 October 2017
91.67% for ultraviolet, 84.21% and 96% for the mid-level data fusion, always had better
performance than partial least square discrimination. Samples from Kunming and
Xishuangbanna were different from others and were distributed with positive and negative
principal component 1. In conclusion, the results could provide fundamental construction for
quality control of P. polyphylla var. yunnanensis for further analysis.
Acknowledgements
This work was supported by the National Natural Science Foundation of China (No.
81460584) and Special Fund for Agro-scientific Research in the Public Interest (No.
201303117).
References
Bergner, N., T. Bocklitz, B. F. M. Romeike, R. Reichart, R. Kalff, C. Krafft, and J. Popp. 2012.
Identification of primary tumors of brain metastases by Raman imaging and support
vector machines. Chemometrics and Intelligent Laboratory Systems 117:224–32.
doi:10.1016/j.chemolab.2012.02.008
Biancolillo, A., R. Bucci, A. L. Magrì, A. D. Magrì, and F. Marini. 2014. Data-fusion for
multiplatform characterization of an Italian craft beer aimed at its authentication.
Analytica Chimica Acta 820:23–31. doi:10.1016/j.aca.2014.02.024
16
Downloaded by [University of Florida] at 21:29 27 October 2017
Casian, T., A. Reznek, A. L. Vonica-Gligor, J. V. Renterghem. 2017. Development, validation
and comparison of near infrared and Raman spectroscopic methods for fast
characterization of tablets with amlodipine and valsartan. Talanta 167:333–43.
doi:10.1016/j.talanta.2017.01.092
Dhanoa, M. S., S. J. Lister, R. Sanderson, and R. J. Barnes. 1994. The link between
multiplicative scatter correction (MSC) and standard normal variate (SNV)
transformations of NIR spectra. Journal of Near Infrared Spectroscopy 2 (1):43–47.
doi:10.1255/jnirs.30
Dupuy, N., O. Galtier, D. Ollivier, P. Vanloot, and J. Artaud. 2010. Comparison between NIR,
MIR, concatenated NIR and MIR analysis and hierarchical PLS model. Application to
virgin olive oil analysis. Analytica Chimica Acta 666 (1–2):23–31.
doi:10.1016/j.aca.2010.03.034
Li, H. 1998. The genus Paris (Trilliaceae). Beijing: Science Press.
Li, Y., J. Zhang, H. Jin, H. Liu, and Y. Wang. 2016. Ultraviolet spectroscopy combined with
ultra-fast liquid chromatography and multivariate statistical analysis for quality
assessment of wild Wolfiporia extensa from different geographical origins.
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 165:61–68.
doi:10.1016/j.saa.2016.04.012
Li, Y., J. Zhang, T. Li, H. Liu, J. Li, and Y. Wang. 2017. Geographical traceability of wild
Boletus edulis based on data fusion of FT-MIR and ICP-AES coupled with data mining
methods (SVM). Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy
177:20–27. doi:10.1016/j.saa.2017.01.029
Lohumi, S., S. Lee, H. Lee, and B. K. Cho. 2015. A review of vibrational spectroscopic
techniques for the detection of food authenticity and adulteration. Trends in Food Science
& Technology 46 (1):85–98. doi:10.1016/j.tifs.2015.08.003
Man, S., W. Gao, Y. Zhang, J. Wang, W. Zhao, L. Huang, and C. Liu. 2010. Qualitative and
quantitative determination of major saponins in Paris and Trillium by HPLC-ELSD and
HPLC-MS/MS.
Journal
of
Chromatography
B
878
(29):2943–48.
doi:10.1016/j.jchromb.2010.08.033
Márquez, C. M. I. López, I. Ruisánchez, and M. P. Callao. 2016. FT-Raman and NIR
spectroscopy data fusion strategy for multivariate qualitative analysis of food fraud.
Talanta 161:80–86. doi:10.1016/j.talanta.2016.08.003
Nohara, T., H. Yabuta, M. Suenobu, R. Hida, K. Miyahara, and T. Kawasaki. 1973. Steroid
glycosides in Paris polyphylla SM. Chemical and Pharmaceutical Bulletin 21
(6):1240–47. doi:10.1248/cpb.21.1240
Qi, L. M., J. Zhang, Y. L. Zhao, Z. T. Zuo, Y. Z. Wang, and H. Jin. 2017. Characterization of
Gentiana rigescens by ultraviolet-visible and infrared spectroscopies with chemometrics.
Analytical Letters 50 (9):1497–511. doi:10.1080/00032719.2016.1225751
17
Downloaded by [University of Florida] at 21:29 27 October 2017
Saptoro, A., M. O. Tadé, and H. Vuthaluru. 2012. A modified Kennard-Stone algorithm for
optimal division of data for developing artificial neural network models. Chemical
Product and Process Modeling 7 (1):1–14. doi:10.1515/1934-2659.1645
Shen, F., D. Yang, Y. Ying, B. Li, Y. Zheng, and T. Jiang. 2012. Discrimination between
Shaoxing wines and other Chinese rice wines by near-infrared spectroscopy and
chemometrics.
Food
and
Bioprocess
Technology
5
(2):786–95.
doi:10.1007/s11947-010-0347-z
Singh, S. B., R. S. Thakur, and H. R. Schulten. 1980. Spirostanol saponins from Paris
polyphylla, structures of polyphyllin C, D, E and F. Phytochemistry 21 (12):2925–29.
doi:10.1016/0031-9422(80)85070-9
Sun, W., X. Zhang, Z. Zhang, and R. Zhu. 2017. Data fusion of near-infrared and mid-infrared
spectra for identification of Rhubarb. Spectrochimica Acta Part A: Molecular and
Biomolecular Spectroscopy 171:72–79. doi:10.1016/j.saa.2016.07.039
Vapnik, V. 2013. The nature of statistical learning Theory. Netherland: Springer Science &
Business Media.
Wu, X., L. Wang, G. C. Wang, H. Wang, Y. Dai, X. X. Yang, W. C. Ye, and Y. L. Li. 2013.
Triterpenoid saponins from rhizomes of Paris polyphylla var. yunnanensis. Carbohydrate
Research 368:1–7. doi:10.1016/j.carres.2012.11.027
Yang, Y., H. Jin, J. Zhang, J. Zhang, and Y. Wang. 2017. Quantitative evaluation and
discrimination of wild Paris polyphylla var. yunnanensis (Franch.) Hand.-Mazz from
three regions of Yunnan Province using UHPLC-UV-MS and UV spectroscopy couple
with partial least squares discriminant analysis. Journal of Natural Medicines 71
(1):148–57. doi:10.1007/s11418-016-1044-7
Yang, Y., X. Liu, W. Li, Y. Jin, Y. Wu, J. Zheng, W. Zhang, and Y. Chen. 2017. Rapid
measurement of epimedin A, epimedin B, epimedin C, icariin, and moisture in Herba
Epimedii using near infrared spectroscopy. Spectrochimica Acta Part A: Molecular and
Biomolecular Spectroscopy 171:351–60. doi:10.1016/j.saa.2016.08.033
Yang, Y., J. Zhang, H. Jin, H. Jin, J. Zhang, and Y. Wang. 2016. Quantitative analysis in
combination with fingerprint technology and chemometric analysis applied for evaluating
six species of wild Paris using UHPLC-UV-MS. Journal of Analytical Methods in
Chemistry 2016:1–9. doi:10.1155/2016/3182796
Yang, Y. G., J. Zhang, Y. L. Zhao, J. Y. Zhang, and Y. Z. Wang. 2017. Quantitative
determination and evaluation of Paris polyphylla var. yunnanensis with different
harvesting times using UPLC-UV-MS and FT-IR spectroscopy in combination with
partial least squares discriminant analysis. Biomedical Chromatography 31 (7):e3913.
doi:10.1002/bmc.3913
Yao, S., T. Li, H. G. Liu, J. Q. Li, and Y. Z. Wang. 2017. Geographic characterization of
Leccinum rugosiceps by ultraviolet and infrared spectral fusion. Analytical Letters 50
(14):2257–69. doi:10.1080/00032719.2017.1279172
18
Downloaded by [University of Florida] at 21:29 27 October 2017
Yin, X., C. Qu, Z. Li, Y. Zhai, S. Cao, L. Lin, L. Feng, L. Yan, and J. Ni. 2013. Simultaneous
determination and pharmacokinetic study of polyphyllin I, polyphyllin II, polyphyllin VI
and polyphyllin VII in beagle dog plasma after oral administration of Rhizoma Paridis
extracts
by
LC-MS-MS.
Biomedical
Chromatography
27
(3):343–48.
doi:10.1002/bmc.2797
Yu, X., Q. Wu, W. Lu, Y. Wang, X. Ma, Z. Chen, and C. Yan. 2013. Metabolomics study of
lung cancer cells based on liquid chromatography-mass spectrometry. Chinese Journal of
Chromatography 31 (7):691–6. doi:10.3724/SP.J.1123.2012.12039
Zhang, J., Y. Wang, Y. Zhao, S. Yang, J. Zhang, T. Yuan, J. Wang, and H. Jin. 2012. Ultraviolet
absorption spectrum analysis and identification of medicinal plants of Paris.
Spectroscopy
and
Spectral
Analysis
32
(8):2176–80.
doi:10.3964/j.issn.1000-0593(2012)08-2176-05
Zhang, T., H. Liu, X. T. Liu, D. R. Xu, X. Q. Chen, and Q. Wang. 2010. Qualitative and
quantitative analysis of steroidal saponins in crude extracts from Paris polyphylla var.
yunnanensis and P. polyphylla var. chinensis by high performance liquid chromatography
coupled with mass spectrometry. Journal of Pharmaceutical and Biomedical Analysis 51
(1):114–24. doi:10.1016/j.jpba.2009.08.020
Zhang, Y., Q. Cong, Y. Xie, J. Yang, and B. Zhao. 2008. Quantitative analysis of routine
chemical constituents in tobacco by near-infrared spectroscopy and support vector
machine. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 71
(4):1408–13. doi:10.1016/j.saa.2008.04.020
Zhao, Y., J. Zhang, T. Yuan, T. Shen, Y. Hou, S. Yang, W. Li, Y. Wang, and H. Jin. 2014. Study
on rapid identification of medicinal plants of Paris polyphylla from different origin areas
by NIR spectroscopy. Spectroscopy and Spectral Analysis 34 (7):1831–35.
doi:10.3964/j.issn.1000-0593(2014)07-1831-05
Zhao, Y., J. Zhang, T. Yuan, T. Shen, W. Li, S. Yang, Y. Hou, Y. Wang, and H. Jin. 2014.
Discrimination of wild Paris based on near infrared spectroscopy and high performance
liquid chromatography combined with multivariate analysis. PLoS One 9 (2):e89100.
doi:10.1371/journal.pone.0089100
Zimmermann, B., and A. Kohler. 2013. Optimizing Savitzky-Golay parameters for improving
spectral resolution and quantification in infrared spectroscopy. Applied Spectroscopy 67
(8):892–902. doi:10.1366/12-06723
19
Table 1. Detailed information of the geographical origin of P. polyphylla var. yunnanensis
Species
Site
Num
Latitude and Longitude
ber
P. polyphylla var.
Kunming City
7
Downloaded by [University of Florida] at 21:29 27 October 2017
yunnanensis
P. polyphylla var.
Kunming City
7
yunnanensis
P. polyphylla var.
Nujiang of Lisu Autonomous
yunnanensis
Prefecture
P. polyphylla var.
Yuxi City
22
19
yunnanensis
P. polyphylla var.
Xishuangbanna of Dai Autonomous
yunnanensis
Prefecture
P. polyphylla var.
Wenshan of Zhuang and Miao
yunnanensis
Autonomous Prefecture
P. polyphylla var.
Wenshan of Zhuang and Miao
yunnanensis
Autonomous Prefecture
20
14
9
4
N25°45′4
E102°31′5
6.07″
0.24″
N25°11′2
E102°50′4
4.74″
4.45″
N25°49′5
E99°03′23.
3.48″
34″
N24°55′4
E102°07′2
6.06″
2.99″
N21°59′1
E100°15′2
1.11″
9.34″
N23°45′0
E104°16′4
2.51″
4.60″
N22°51′2
E104°6′11.
4.41″
83″
Table 2. Best C, best gamma, best cross validation accuracy rate, root mean square error of
estimation and root mean square error of cross validation in the support vector machine and
partial least square discriminant analysis for Fourier transform infrared and ultraviolet
spectroscopy
Support vector machine
Partial least square
Downloaded by [University of Florida] at 21:29 27 October 2017
discriminantion analysis
Be Best
Best cross
Root mean
Root mean
st
gam
validation
square error
square error of
C
ma
accuracy
of estimation
cross
rate
Origin
Fourier
11
1.66
transfor
41
×
m
.0
e6
infrared
4
Ultraviol
27
0.00
et
.8
13
validation
57.89%
0.3912
0.4095
60.34%
0.2966
0.3085
6
Multiplicative scatter
Fourier
1
0.1
0%
0.3932
0.3979
correction
transfor
Ultraviol
1.
0.02
64.91%
0.2914
0.3123
et
74
Fourier
2
0.1
26.32%
0.3931
0.3980
m
infrared
Standard normal variate
21
transfor
m
infrared
Ultraviol
44
0.00
et
5.
68
64.91%
0.3091
0.3132
64.91%
0.3610
0.4323
73.68%
0.2821
0.3218
70.18%
0.3505
0.3966
77.59%
0.2536
0.3037
Downloaded by [University of Florida] at 21:29 27 October 2017
72
Savitzky-Golay(11) +
Fourier
second derivative
transfor
16
0.00
04
m
infrared
Ultraviol
84
0.00
et
.4
22
5
Savitzky-Golay(11) +
Fourier
second derivative +
transfor
standard normal variate
m
2
0.00
28
infrared
Ultraviol
et
16
0.00
68
22
Table 3. Parameter C and gamma, accuracy rate of calibration and validation for Fourier
transform infrared spectroscopy and ultraviolet spectroscopy with low- and mid-level data fusion
Support vector machine
Partial least square
discrimination
Downloaded by [University of Florida] at 21:29 27 October 2017
Best C
Fourier
Best
Accuracy
Accuracy
Accuracy rate
Accuracy
gamma
rate of
rate of
of calibration
rate of
calibration
validation
validation
2
0.0027
71.93%
92%
22.81%
20%
Ultraviolet
32
0.0055
77.59%
91.67%
70.18%
68%
Low-data
1448.15
2.69 × e6
84.21%
92%
40.35%
80%
2.8284
0.0884
84.21%
96%
96.49%
84%
transform
infrared
fusion
Mid-data
fusion
23
Downloaded by [University of Florida] at 21:29 27 October 2017
Figure 1. Unprocessed ultraviolet spectra.
24
Figure 2. Preprocessing based on second derivative, standard normal variation and
Downloaded by [University of Florida] at 21:29 27 October 2017
Savitzky-Golay (11 points) smoothed the ultraviolet spectra.
25
Downloaded by [University of Florida] at 21:29 27 October 2017
Figure 3. Unprocessed Fourier transform infrared spectra.
26
Figure 4. Preprocessing based on second derivative, standard normal variation and
Downloaded by [University of Florida] at 21:29 27 October 2017
Savitzky-Golay (11 points) smoothed Fourier transform infrared spectra.
27
Downloaded by [University of Florida] at 21:29 27 October 2017
Figure 5. Parameters C and gamma obtained with mid-level data fusion.
28
Downloaded by [University of Florida] at 21:29 27 October 2017
Figure 6. Accuracy of the validation set by mid-level data fusion.
29
Figure 7. Two-dimension score plot of partial least square discriminant analysis for P.
Downloaded by [University of Florida] at 21:29 27 October 2017
polyphylla var. yunnanensis from different geographical origins.
30
Документ
Категория
Без категории
Просмотров
0
Размер файла
1 112 Кб
Теги
2017, 1385618, 00032719
1/--страниц
Пожаловаться на содержимое документа