Abstract
Liquid chromatographychemometric methods [LC-Partial least squares (LC-PLS), LC-principle
component regression (LC-PCR) and LC-artificial neural network (LC-ANN)] were developed for
the determination of ( )-anomalin (ANO) and deltoin (DEL) in the root and aerial part of Seseli
resinosum Freyn et Sint. (Umbelliferae). Firstly, chemometric conditions were optimized by testing
different mobile phases at various proportions of solvents with various flow rates in different
wavelengths by using a normal phase column to obtain the best separation and recovery results.
As a result, a mobile phase consisting of n-hexane and ethyl acetate (75:25 v/v) at a constant flow
rate of 0.8 mL min 1 on the above column system at ambient temperature were found to be the
optimal chromatographic condition for good separation and determination of ANO and DEL in
samples. Multichromatograms for the concentration set containing ANO and DEL compounds in
the concentration range of 50400 ng mL 1 were obtained by using a diode array detector
(DAD) system at selected wavelength sets, 300 (A), 310 (B), 320 (C), 330 (D) and 340 (E). Three
LC-chemometric approaches were applied to the multichromatographic data to construct chemometric calibrations. As an alternative method, traditional LC at single wavelength was used for
the analysis of the related compounds in the plant extracts. All of the methods were validated by
analyzing various synthetic ANODEL mixtures. After the above step, traditional and chemometric LC methods were applied to the real samples consisting of extracts from roots and aerial
parts of S. resinosum. The results obtained by LC-chemometric approaches were compared to
each other and with those obtained by traditional LC.
Keywords
Column liquid chromatography
Multichromatographic data
Chemometric-LC calibration
( )-Anomalin and deltoin
Introduction
The genus Seseli L. is represented by 12
taxa (11 species and one subspecies) in
Original
DOI: 10.1365/s10337-007-0392-6
0009-5893/07/11
found in Seseli genus have some pharmacological eects. Their most important
activities are anticoagulant, antitumoral,
immunomodulating, and antiinammatory eects [5]. ( )-Anomalin (ANO)
and deltoin (DEL) in our study are
categorized into coumarin derivatives
isolated from the n-hexane extract of
the root of S. resinosum [6]. The molecular structure of ANO and DEL corresponding to (I) and (II), are shown in
Fig. 1.
No determination of ANO and DEL
in the same mixture was observed in the
literature until now. For the above reasons, the quantitative analysis of the
investigated ANO and DEL is a very
important task for future studies.
Generally, a traditional LC method
has been used for the quality control of
commercial products and their routine
analysis in laboratory, and for the determination of some active compounds in
plant extract samples. However, new
combined analytical methods called
hyphenated instrumentations such as LC
DAD and LCMS, etc., have been used
for the quantitative analysis of two or
more active compounds in complex mixtures. In some cases, the above modern
analytical methods may not give desirable
results for the multicomponent determination of compounds in samples due to
interference and matrix eects or other
environmental factors on the analysis.
Taking into account all above arguments,
the analysis of complex mixtures needs
new analytical approaches based on the
677
Experimental
LC-Chemometric Methods
In this investigation, the multichromatograms of ANO and DEL at the ve
wavelengths via DAD system are plotted
and stored in a computer. The detector
responses at the multiwavelengths are
measured in terms of peak area. The recorded multichromatographic data corresponding to the concentration and
sample sets are transferred into Microsoft
EXCEL and are processed by three different chemometric calibration methods,
PLS, PCR and ANN mathematical
algorithms. The theoretical details of the
application of the chemometric approaches to the multichromatographic
data are explained below.
678
LC-PLS Method
LC-ANN Method
Instrumentation
and Chromatography
A Shimadzu UV-160 double beam UV
Vis spectrophotometer possessing a xed
slit width (2 nm) connected to a computer loaded with Shimadzu UVPC
software was used to record the absorption spectra.
Chromatography was performed with
an Agilent 1100 series HPLC system (Agilent Technologies, CA, USA) provided
with a quaternary pump, a thermostatted
autosampler, a thermostatted column
compartment, and a multiwavelength
diode array detector (DAD). Data
were acquired and processed using HP
Chem Station for LC [Rev. A0.01 (403)]
(Agilent). Waters Spherisorb S5W
(25 cm 4 mm 5 lm) column was used
for the chromatographic separation. The
ow rate was maintained at 0.8 mL min 1
and the injection volume was 10 lL. The
mobile phase was prepared daily and ltered through a 0.45 lm membrane lter.
In fact the same chromatographic
conditions were used for both, traditional
and LC chemometric analyses.
Standard Solutions
Stock solution of ANO and DEL
(50 mg 50 mL 1 for each compound)
were prepared with n-hexane and ethyl
acetate (50:50 v/v). A concentration set of
mixture solutions containing both compounds in the range of 50400 ng mL 1
was obtained from the above stock solutions. A validation set consisting of six
synthetic mixture solutions of ANO and
DEL in the above concentration range
was prepared. All the solutions were
prepared freshly and protected from
light.
Original
Sample Collection
and Preparation for Analysis
O
O
O
O
(I)
(II)
1.6
1.4
1.2
Abs.
1.0
0.8
0.6
0.4
0.2
0.0
240
260
280
300
320
340
360
380
Wavelength (nm)
Fig. 2. Absorption spectra of ANO () and DEL () compounds in the mixture of n-hexane and
ethyl acetate (50:50 v/v)
679
mAU
200
A, 1 =300 nm
B, 2 =31 0 nm
C, 3 =320 nm
D, 4 =330 nm
E, 5 =340 nm
175
150
mAU
A, 1 =300 nm
125
B, 2 =310 nm
100
75
50
25
C, 3 =320 nm
1600
D, 4 =330 nm
B
A
0
0
10
15
Concentration set 1
E, 5 =340 nm
25
20
Time (min)
mAU
A, 1 =300 nm
B, 2 =31 0 nm
C, 3 =320 nm
D, 4 =330 nm
E, 5 =340 nm
600
500
400
1400
300
E
D
200
1200
C
100
B
A
0
0
10
15
Concentration set 2
20
25
Time (min)
mAU
A, 1 =300 nm
B, 2 =31 0 nm
C, 3 =320 nm
D, 4 =330 nm
E, 5 =340 nm
1200
1000
1000
a
800
600
E
D
400
800
200
B
A
0
0
10
15
Concentration set 3
20
25
Time (min)
mAU
600
A, 1 =300 nm
B, 2 =31 0 nm
C, 3 =320 nm
D, 4 =330 nm
E, 5 =340 nm
1600
1400
1200
1000
800
600
400
400
C
B
200
0
0
10
15
Concentration set 4
20
25
Time (min)
200
A, 1 =300 nm
B, 2 =31 0 nm
C, 3 =320 nm
D, 4 =330 nm
E, 5 =340 nm
mAU
2000
1750
1500
1250
1000
750
500
250
Concentration set 4
B
A
0
0
10
15
Time (min)
Concentration set 5
20
25
10
15
20
25
Time (min)
Multichromatographic Data
Processing
A concentration set of ve dierent mixture solutions of ANO and DEL in the
680
ANO a and DEL b in the concentration set 4. Small multiple chromatograms correspond to the
Original
Table 1. Concentration set and its corresponding normalized peak area data
Concentration set
1
ANO (ng mL )
DEL (ng mL )
300 nm
310 nm
320 nm
330 nm
340 nm
300 nm
310 nm
320 nm
330 nm
340 nm
50
100
200
300
400
50
100
200
300
400
0.1591
0.2873
0.4832
0.6991
1.0000
0.1618
0.3135
0.5260
0.8031
0.9694
0.1739
0.3109
0.5217
0.7957
0.9647
0.1775
0.3160
0.5296
0.8069
0.9779
0.1648
0.3069
0.5139
0.7824
0.9447
0.1661
0.2773
0.5832
0.7476
0.9799
0.1680
0.2860
0.6006
0.7683
0.9953
0.1725
0.2873
0.6029
0.7707
0.9974
0.1708
0.2841
0.5956
0.7602
0.9809
0.1751
0.2900
0.6073
0.7747
1.0000
300 nm
310 nm
ANO
m
n
r
SE(n)
SE(m)
SE(r)
LOD (ng mL 1)
LOQ (ng mL 1)
DEL
2.33 10
3.65 10
0.9966
2.73 10
1.11 10
3.18 10
5.66
18.86
3
2
2
4
2
2.32 10
6.28 10
0.9956
3.09 10
1.26 10
3.60 10
4.07
13.57
320 nm
ANO
3
2
2
4
2
DEL
2.33 10
6.61 10
0.9967
2.69 10
1.09 10
3.13 10
7.84
26.14
3
2
2
4
2
2.36 10
6.71 10
0.9950
3.39 10
1.38 10
3.94 10
5.04
16.79
330 nm
ANO
3
2
2
4
2
DEL
2.29 10
7.28 10
0.9974
2.33 10
9.49 10
2.72 10
8.36
27.85
3
2
2
5
2
2.36 10
7.03 10
0.9950
3.37 10
1.37 10
3.93 10
11.31
37.69
340 nm
ANO
3
2
2
4
2
DEL
2.32 10
7.51 10
0.9974
2.36 10
9.59 10
2.75 10
9.02
30.06
3
2
2
5
2
ANO
2.32 10
7.11 10
0.9948
3.39 10
1.38 10
3.94 10
9.71
32.38
3
2
2
4
2
2.25 10
6.97 10
0.9970
2.49 10
1.01 10
2.90 10
6.43
21.45
DEL
3
2
2
4
2
2.36 10
7.33 10
0.9948
3.44 10
1.40 10
4.01 10
10.17
33.89
3
2
2
4
2
Table 3. Recovery data obtained by application of the proposed methods to the synthetic mixtures
Mixture
(ng mL 1)
Traditional HPLC
300 nm
ANO
50
200
400
200
200
200
DEL
200
200
200
50
200
400
ANO
Found
45.6
187.8
387.7
189.4
193.1
189.2
310 nm
DEL
(ng mL
191.1
192.9
193.2
46.4
194.7
403.3
Recovery (%)
91.1
95.6
93.9
96.5
96.9
96.6
94.7
92.8
96.6
97.4
94.6 100.8
Mean
SD
RSD (%)
94.6
2.10
2.21
ANN
96.6
2.61
2.70
ANO
320 nm
330 nm
PCR
PLS
340 nm
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
)
45.7
198.8
401.3
195.2
195.0
203.6
197.4
188.1
194.0
48.2
184.7
394.5
46.5
197.7
405.1
183.4
193.4
188.0
183.0
186.1
195.5
48.6
184.6
393.7
44.5
198.5
399.2
190.7
193.6
196.2
198.2
194.2
201.8
45.7
203.2
400.3
45.0
202.5
413.2
186.5
194.7
198.2
191.6
189.7
189.2
51.2
182.8
376.8
46.6
198.3
393.8
198.5
200.3
199.8
195.0
193.9
199.6
51.7
196.7
400.0
47.9
197.0
400.7
189.2
194.0
195.1
192.4
190.4
194.9
49.6
190.2
391.9
48.0
197.0
400.7
189.2
194.0
195.1
192.4
190.9
195.1
49.7
190.2
392.0
91.3
99.4
100.3
97.6
97.5
101.8
98.7
94.1
97.0
96.3
92.3
98.6
93.0
98.9
101.3
91.7
96.7
94.0
91.5
93.0
97.8
97.1
92.3
98.4
89.1
99.2
99.8
95.3
96.8
98.1
99.1
97.1
100.9
91.4
101.6
100.1
89.9
101.2
103.3
93.2
97.4
99.1
95.8
94.9
94.6
102.5
91.4
94.2
93.3
99.2
98.5
99.2
100.1
99.9
97.5
97.0
99.8
103.3
98.3
100.0
95.7
98.5
100.2
94.6
97.0
97.5
96.2
95.2
97.4
99.2
95.1
98.0
96.0
98.5
100.2
94.6
97.0
97.5
96.2
95.5
97.5
99.4
95.1
98.0
98.0
3.66
3.73
96.2
2.54
2.64
95.9
3.69
3.85
95.0
3.07
3.23
96.4
3.93
4.08
98.4
3.75
3.81
97.4
5.00
5.14
95.6
3.70
3.88
98.4
2.56
2.60
99.3
2.31
2.33
97.3
1.99
2.04
96.9
1.64
1.69
97.3
1.95
2.00
97.0
1.65
1.70
LC-PLS Method
In the LC-PCR method, the concentration set and normalized chromatographic data sets given in Table 1 were
considered for LC-PLS calibration. In
all of the chemometric and traditional
calibration models, normalized chromatographic peak area values were used.
According to the PLS algorithm, LCPLS calibration were computed by using
the linear combination based on the
Original
LC-PCR Method
681
Table 4. Assay results obtained by applying the proposed analytical approaches to the root
samples
Method
Root samples
Traditional 300
HPLC
310
320
330
340
ANN
PCR
PLS
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
ANO
DEL
Mean
SD
(lg mg 1,
%)
2.06
2.95
2.01
2.88
2.15
3.07
2.09
2.96
2.16
3.03
2.16
3.01
2.10
2.98
2.09
2.99
2.08
2.96
2.04
2.92
2.10
2.95
2.08
2.95
2.10
3.10
2.17
3.00
2.10
2.97
2.09
2.98
2.07
2.98
2.07
3.04
2.09
3.08
2.18
3.08
2.07
3.07
2.20
3.07
2.12
3.02
2.11
3.01
2.12
2.95
2.07
2.88
2.10
2.93
2.06
2.91
2.07
3.06
2.19
3.04
2.13
2.97
2.09
3.01
2.04
3.01
2.07
2.95
2.12
2.99
2.11
2.99
2.13
2.98
2.17
3.00
2.10
2.99
2.11
3.00
2.04
2.98
2.06
2.89
2.08
3.05
2.08
2.94
2.04
3.09
2.14
3.01
2.07
2.99
2.08
3.00
2.04
3.03
2.01
2.87
2.15
2.91
2.15
3.04
2.10
3.00
2.19
3.00
2.10
2.97
2.09
2.98
2.06
2.98
2.05
2.92
2.11
3.00
2.11
2.98
2.10
3.05
2.17
3.02
2.10
2.98
2.09
3.00
0.03
0.03
0.03
0.06
0.03
0.07
0.04
0.06
0.04
0.05
0.02
0.03
0.02
0.02
0.01
0.01
RSD
1.56
1.02
1.40
2.07
1.36
2.30
2.11
1.97
1.92
1.50
0.96
0.93
0.89
0.63
0.55
0.39
were used to calculate the LC-ANN calibration. In this case, normalized chromatographic data for the concentration
data set (see Table 1) were utilized as an
input data. To reach the optimal HPLCANN calibration model with dierent
neuron size, various topological networks
were assayed and a training network
having ve neurons in the input layer
with one hidden layer and one output for
the prediction of each ANO and DEL in
samples were found to be suitable for the
construction of the LC-ANN calibrations. In this chemometric ANN approach, logarithmic signal transfer
function in three layers was used for
obtaining the LC-ANN calibrations for
each compound.
During a back propagation ANN
algorithm application, the relationship
between the mean square error (MSE) of
the networks and epochs for the training
set was performed. A dierence between
the calculated MSE and predened MSE
(1.0 10 10) values was not observed for
the calculated ANN calibration. The obtained LC-ANN calibration was used for
the prediction of ANO and DEL in
samples mentioned above. ANN application gives us a good agreement with
PCR and PLS calibrations.
Traditional LC Method
682
Original
Method Validation
The validity of the LC-chemometric and
traditional LC methods was performed
for obtaining reliable results of analysis.
This procedure was carried out by using
the following parameters. In the traditional LC method, linearity for ANO and
DEL was observed in the concentration
range of 50400 ng mL 1 because the
values of correlation coecients being
close to 1.0 as depicted in Table 2. In case
of the traditional LC method, the limit of
detection (LOD) and the limit of quantitation (LOQ) were calculated by using
100 ng mL 1 for ANO and DEL for six
replicates and their numeric values were
presented in Table 2.
The percent of mean recoveries and
their relative standards deviations for two
compounds were obtained by application
of all of the proposed methods to six
dierent mixtures of ANO and DEL in
dierent concentration levels. Accuracy
and precision corresponding to the percent of mean recoveries and relative
standards are shown in Table 3. We observed that the proposed methods are
available for the quantitative analysis of
the subjected compounds due to the
above-mentioned accuracy and precision
values.
Sample Analysis
In the application of the proposed
methods to the real samples, LC-ANN,
LC-PCR, LC-PLS and traditional LC
approaches were successfully applied to
the quantitative analysis of ANO and
DEL in the root and aerial part of S.
resinosum. As an example for the chromatographic separation obtained in the
method application to the sample, a
multiple chromatogram of the root extract is presented in Fig. 4.
Assay results (n = 7) obtained in the
application of the proposed traditional
LC and LC-chemometric approaches to
the root and aerial part samples are given
in Tables 4 and 5, respectively. Generally
successful results provided by the LCchemometric and traditional LC methods
were obtained for the analysis of related
compounds. Therefore, we observed that
the ANO and DEL % RSD values
Original
Table 5. Assay results obtained by applying the proposed analytical approaches to the aerial part
samples
Method
SD RSD
Mean
(lg mg 1,
%)
4.92
11.06
4.48
10.94
4.36
10.77
4.37
10.65
4.44
10.52
4.58
10.11
4.43
10.76
4.51
10.79
4.78
10.90
4.14
10.26
4.14
10.08
4.16
9.95
4.23
10.23
4.38
10.21
4.50
10.46
4.51
10.49
4.71
11.23
4.54
11.11
4.41
10.85
4.43
10.78
4.49
10.65
4.50
10.19
4.53
10.80
4.53
10.72
4.73
11.23
4.52
11.08
4.39
10.88
4.41
10.75
4.47
10.62
4.50
10.51
4.52
10.88
4.51
10.90
4.69
10.72
4.28
10.57
4.15
10.74
4.15
10.25
4.23
10.13
4.36
10.48
4.41
10.56
4.43
10.58
4.61
11.03
4.31
10.72
4.19
11.04
4.22
10.14
4.31
10.07
4.47
10.28
4.44
10.58
4.51
10.61
4.61
10.91
4.64
10.68
4.39
11.03
4.38
10.50
4.46
10.41
4.45
10.40
4.41
10.68
4.50
10.69
4.72
11.01
4.42
10.77
4.29
10.77
4.30
10.43
4.38
10.38
4.46
10.31
4.46
10.68
4.50
10.68
0.11
0.19
0.18
0.30
0.12
0.33
0.12
0.32
0.11
0.23
0.07
0.15
0.05
0.15
0.03
0.14
2.28
1.69
3.98
2.81
2.88
3.04
2.84
3.10
2.60
2.25
1.64
1.49
1.16
1.40
0.70
1.31
Conclusions
References
683