Application quantum and physico chemical molecular descriptors utilizing principal components to study mode of anticoagulant activity of pyridyl chromen-2-one derivatives

Article abstract of DOI:10.1016/j.bmc.2008.12.055

Factors II, V, VII and Xa have materialized as a key enzymes for the intervention of blood coagulation cascade and for the development of new anti thrombotic agents. The combined density functional quantum mechanical/molecular mechanical (QM/MM) approach has been used to access inhibition of prothrombin and thrombin production. The biological activities of coumarin derivatives as clotting factor inhibitors was quantitatively analyzed in terms of physicochemical parameters utilizing the principal component analysis. Structural requirements for maximal potency were derived from the results of a quantitative structure activity relationship analysis.

Full text of DOI:10.1016/j.bmc.2008.12.055

Bioorganic & Medicinal Chemistry 17 (2009) 1654–1662
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Application quantum and physico chemical molecular descriptors utilizing
principal components to study mode of anticoagulant activity of pyridyl
chromen-2-one derivatives
a
a
a
b
M. S. Bhatia a, , K. B. Ingale , P. B. Choudhari , N. M. Bhatia , R. L. Sawant
*
a Department of Pharmaceutical Chemistry, Bharati Vidyapeeth College of Pharmacy, Near Chitranagri, Kolhapur-416013 (MS), India
b Department of Pharmaceutical Chemistry, PDVVF‘s College of Pharmacy, Ahemadnagar – 414111 (MS), India
a r t i c l e i n f o
a b s t r a c t
Article history:
Factors II, V, VII and Xa have materialized as a key enzymes for the intervention of blood coagulation cas-
cade and for the development of new anti thrombotic agents. The combined density functional quantum
mechanical/molecular mechanical (QM/MM) approach has been used to access inhibition of prothrombin
and thrombin production. The biological activities of coumarin derivatives as clotting factor inhibitors
was quantitatively analyzed in terms of physicochemical parameters utilizing the principal component
analysis. Structural requirements for maximal potency were derived from the results of a quantitative
structure activity relationship analysis.
Received 15 October 2008
Revised 20 December 2008
Accepted 23 December 2008
Available online 29 December 2008
Keywords:
Prothrombin time
Activated partial thromboplastin time
Chromen-2-one
Ó 2008 Elsevier Ltd. All rights reserved.
QSAR
1. Introduction
forming (prophylaxis) and getting larger.3,4 The main side effect
associated with the use of coumarin derivatives is hemorrhaging.
Anticoagulant agents are used to prevent the extension of blood
clots in phlebitis or deep vein thrombosis and as a prophylactic
agent in patients that have mechanical heart valves (life-long ther-
apy). Fatal damage is often caused by thrombotic events. Hence,
numerous efforts have been made to synthesize anti thrombotics,
such as antiplatelet, anticoagulant, and thrombolytic agents. Pro-
thrombin and thrombin plays a critical role in thrombosis, since
they not only convert fibrinogen to fibrin for clot formation but also
strongly induces platelet aggregation. The human factor Xa is a tryp-
sin-like serine protease that serves a critical role in blood coagula-
tion events.1 The factor Xa combines with factor Va and platelet
phospholipid, as a result a prothrombinase complex is formed. This
prothrombinase complex, in turn, initiates the splitting of pro-
thrombin to form thrombin thereby setting in motion the final clot-
ting process. Since it was calculated that a molecule of factor Xa
could generate 138 molecules of thrombin,2 inhibition of factor Xa
may be more efficient than inactivation of thrombin in interrupting
the blood coagulation system. Coumarin ring containing products
inhibit coagulation by interfering with the incorporation of vitamin
K into vitamin-K dependent clotting factors (Factors II, VII, IX, and
X). Coumarin products do not dissolve clots but prevent clots from
The degree of anticoagulation is measured by a blood sample and
expressed as the International Normalized Ratio (INR). The INR is
the international standard. The therapeutic INR is generally be-
tween 2 and 3.5 which correlates with a 30–50% inhibition of vita-
min-K dependent clotting factors. Ideally, the patient should have
his/her INR checked every 4–6 weeks while on this medication.
The main objective of the present report was to use quantum
and other physicochemical molecular descriptors (MDs) to gener-
ate predictive linear discriminant analysis (LDA)-assisted QSAR
models, enabling the selection of novel coumarin derivatives with
good anticoagulant activity. The data sets were generated by syn-
thesizing various coumarin derivatives and testing them for elon-
gation of prothrombin time (PT) and activate partial thrombin
time (APTT). Prothrombin time was measured by utilizing rabbit
brain thromboplastin reagent and measuring the time required
for the formation of clot. In order to determine the Activated par-
tial thromboplastin time brain extract was incubated with Kaolin
(Factor XII) and the clotting time in seconds was measured after
addition of calcium ions.
The synthetic availability and interesting structural features of
chromen-2-one a ubiquitous structural unit in many biologically
active compounds prompted us to carry out an extension toward
the preparation of 3-(6-aminopyridin-3-yl)-4-methyl-2H-chro-
men-2-one derivatives (Fig. 1) for synthesis and QSAR studies of
4-methyl-3-(6-{[arylmethylene] amino} pyridine-3-yl)-2H-chro-
* Corresponding author. Tel.: +91 9822172940.
E-mail address: manish_bhatia13@yahoo.co.in (M.S. Bhatia).
URL: http://drmanishbhatia.com (M.S. Bhatia).
0968-0896/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2008.12.055

M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
1655
men-2-one as anti-coagulant agents. The compounds obtained by
substitutions are shown in Table 1.
2.2. Molecular diversity study and design of the training and
test sets using CA
The quality of a classification (or any QSAR) model depends on
the quality of the selected dataset. One of the most critical aspects
of constructing the training set is to warrant enough molecular
diversity for it. Taking this into account, we selected a database of
64 compounds having a great degree of structural variability. To
demonstrate the structural diversity of these datasets, we per-
formed hierarchical CAs of the series.12–16 The dendrograms are gi-
ven in Figures 2 and 3, using the Euclidean distance (X-axis) and
the complete linkage (Y-axis), illustrating the results of the k-NNCA
developed into active and inactive sets, correspondingly. As can be
seen in the binary tree, there is a great number of different subsets,
which prove the molecular variability of the selected chemicals in
these databases. Because of the difficulty in evaluating the output
dendrogram, other kind of CAs is usually performed. Therefore, to
split the whole group into two data sets (training and test ones),
we perform a k-MCA. The main idea of this procedure consists in
making a partition of anticoagulant series of chemicals into several
statistically representative classes of compounds.12–16 Afterward,
the selection of the training and prediction sets was performed by
taking, in a random way, compounds belonging to each cluster. From
these 64 compounds, 52 were chosen at random to form the training
set. By using a training data set including maximum possible struc-
tural variations makes it possible to explore all structures likely to
have the targeted activity. Thus the discovery of lead compounds
with good anticoagulant activity, not only with determined action
modes but also with novel molecular scaffolds becomes more likely.
The remaining series of 12 compounds was prepared as the test sets
for the external validation of the models. These compounds were
never used in the development of the classification models. The
compounds used in training Set and test set are shown in Table 1.
2. Results
2.1. Analysis of principal components to detect structure
in the relationships between variables (MDs)
To conduct the analysis of information contents and linear
independence of the MDs calculated in this study, we performed
a factorial analysis.5–11 Factorial loadings from the principal com-
ponent analysis, after a varimax normalized rotation of the first
twelve factors, are shown in Table 2. Furthermore, the results
of the factorial analysis performed here are summarized in Table
3. The twelve principal factors explain approximately 99.5% of
the variance. The first factor explains 42.1% of the variance in
the MDs studied. The addition of the second factor increases to
55.9% the explained variance, and the addition of the third factor
allow 70.7% of the index variance to be accounted for. The other
factors are summarized in Table 3. Most of the steric MDs are
strongly loaded into factor 1 (F1), such as Molwt, Vol, SMR, Idav-
erage, MIz, RGYR, Kappa1 and kappa2. Some of these MDs, in
addition to hydrophobic information, also have steric and topo-
logical information. For instance, distTop is also robustly loaded
(loading > 0.70, but with opposite relation) into this factor. Thus,
these ‘F1-MDs’ produce much redundancy and overlapping be-
tween them, and their relations are rather complex. This factor
F1 can be taken into account as a hydrophobic-steric factor.
The second factor (F2) also shows an electronic dimension (DM,
LUMO, QMDM) along with hydrophobic dimensions (H-acc)
strongly loaded. This result showed that electronic and hydro-
phobic MDs have a parallel relation. The third factor (F3) appears
to be most significant for the other hydrophobic features (H-don,
SAaverage). The most of the local charge MDs are strongly loaded
in factors F4 (HF), F5 [IP, HOMO], F6 (QMx), 7 [zdipole], F8
(QMy), showing a rather complex relation. Finally, in factor 9,
10, 11 no MDs are strongly loaded, but the highest loadings were
found to be of electronic parameters (not given in Table 1) there-
fore, these dimension are electronic factors. As stated in Material
and Methods section, these MDs with high loadings in the same
factor are interrelated, while no correlation exists between those
MDs having nonzero loadings only in different factors. Thus, we
can say that the MDs included in each dimension contain struc-
tural information, not contained in any other MD incorporated in
others factors.
2.3. Development of the discriminant function
Although the number of existing statistical methods of getting
classification functions is relatively extensive, we select LDA given
the simplicity of the method.17 The use of LDA in drug design has
been extensively reported by different authors. Therefore, LDA
was also the technique used in the generation of discriminant
functions in the current work. Making use of the LDA technique
implemented in the STATISTICA software,18 the following linear model
was obtained, in which the quantum and physicochemical MDs
were used as independent variables:
O
O
CH3
Br
C2H5
Br2
Ethylacetoacetate
Glacial acetic
acid
O
H2N
N
N
N
H2N
H2N
Substituted
Phenols,
H2SO4
O
O
R
R
R1-CHO
R1
CH3
CH3
1-8
Glacial acetic
acid
N
N
N
H2N
9-64
Figure 1. Figure showing the synthesis of the pyridyl chromen-2-one derivatives.

1656
M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
Table 1
Table 2
Table showing compounds used in this study for the anticoagulant activity
measurements
Factor loadings obtained from the principal component analysis of the quantum and
physicochemical MDs computed in this work
Sr. no
R
X
R1
MDsa
F1
0.98
0.97
F2
F3
F4
F5
F6
F7 F8 F9 F10 F11 F12
1
2
3
4
5
6
7
8
9
—
—
H
NO2
H
NO2
H
NO2
H
NO2
H
H
H
H
—
—
—
—
—
—
—
—
Molwt
Vol
7-OH
7-OH
6-NH2
6-NH2
8-NO2
8-NO2
—
—
—
—
—
—
—
—
—
—
—
—
—
H-acc
H-don
SlogP
SMR
c3c
cV3c
DM
Q1
0.63 0.71
0.70
0.80
0.97
0.85
0.87
ꢀ0.70
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
Q2
IDAverage 0.91
IP
HOMO
LUMO
QMx
QMy
QMz
SKaverage 0.67
SAaverage
MIx
H
H
H
H
ꢀ0.52
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
0.52
ꢀ0.78
NO2
NO2
NO2
NO2
NO2
NO2
NO2
H
H
H
H
H
ꢀ0.62
0.60
H
ꢀ0.73
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
—
MIy
MIz
0.88
0.89
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
7-OH
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
6-NH2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
8-NO2
Xdipole
Ydipole
Zdipole
RGYR
HF
DistTop
K1
K2
ꢀ0.66
0.63
H
0.93
H
H
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
ꢀ0.67
ꢀ0.72
0.97
0.97
NO2
NO2
NO2
NO2
NO2
NO2
NO2
H
H
H
H
H
QMDM
0.79
a
Factor loadings between ꢀ0.5 and 0.5 (0.5 < factor loading > ꢀ0.5) were deleted
H
to clarify the relationships between MDs and identify their factor structure.
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
N = 52, r2 = 0.82, q2 = 0.77, k = 0.96, F = 41.96, pred r2 = 0.78.
pIC50 ðAPTTÞ ¼ 0:0538 þ 0:0013 Heat of Formation
ꢀ 0:0306 H-AcceptorCount
ꢀ 0:0215 XcompDipole
H
H
H
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
NO2
NO2
NO2
NO2
NO2
NO2
NO2
H
H
H
H
H
þ 0:0305 QMDipoleX
ꢀ 0:0299 QMDipoleY
þ 4:0633 SAAverage
ðMODEL 2Þ
H
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
N = 52, r2 = 0.73, q2 = 0.65, k = 0.87, F = 32.25, pred r2 = 0.81.
Where N is the number of compounds, r2 is the correlation coeffi-
cient, q2 is the, k is Wilks’ statistics, F is the Fisher ratio, pred r2 is
the r2 for external test set.. The observed and predicted biological
activities along with the residuals are shown in Tables 4 and 5.
The correlation plot for model 1 and 2 are shown in Figures 4 and
5, respectively.
As we said earlier, the factors obtained by factorial analysis cap-
ture all of the ‘essence’ of the original MDs, because they are linear
combinations of the original items, and these consecutive factors
are uncorrelated to each other. Taken into account this procedure
(the 12 previously extracted factors), the above stated QSAR mod-
els were developed.
H
H
H
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2-Hydroxyphenyl
4-Chlorophenyl
4-Methoxyphenyl
Phenyl
NO2
NO2
NO2
NO2
NO2
NO2
NO2
H
NN-Dimethyl aminophenyl
3,4,5-Trimethoxy phenyl
2.4. Internal and external validations
pIC50ðPTÞ ¼ 2:4769 þ 10:4372 SAAverage
ꢀ 0:0212 H-AcceptorCount
ꢀ 0:3111 IonizationPotential
þ 0:0134 DipoleMoment
Validation of QSAR equations is a very important aspect in li-
gand-based drug design. Together with the quantification of the
goodness-of-fit, it is quite important to obtain a measure of the
model predictive capacity and stability. The most popular valida-
tion criterion to explore the robustness of a predictive model is
through the analysis of the influence of each one of individual ob-
þ 0:0336 QMDipoleX
ꢀ 0:0012 Quadrupole1
ðMODEL 1Þ

M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
1657
Table 3
explains the coumarin-clotting factor interactions. It is mostly
due to the biological activities expressed by the SAAverage, a me-
sure of hydrophobic function, electronic MDs, the Dipole Moment
and quantum mechanic Dipole moment along X -axis, which
showed positive contribution. The positive contribution of SAAver-
age implies that the binding affinity increases with the rise in
hydrophobic features of compounds up to reach a critical value,
after which the affinity will decrease. The electronic parameters
like DM and QMx indicate that electronic interactions are impor-
tant in formation of thrombin. The generated model shows that
biological activity is negatively contributed with MDs like H-
Acceptor count, Ionizationpotential and Quadrapole1. The pres-
ence of Hydrogen bond acceptor moiety in the structure will re-
duce the anticoagulant effect. Thus the addition of hydrogen
bond donor groups like hydroxyl, amino will increase the activity.
Results of the factor analysis by using the principal component method for 23
quantum and physicochemical MDs for 64 compounds
Factor
% Total variancea
% Cumulative variancea
F1
F2
F3
F4
F5
F6
F7
F8
42.1
14.9
11.0
6.1
4.7
3.9
3.6
2.9
2.5
2.1
42.1
55.9
70.8
76.8
81.5
85.4
88.9
91.8
94.2
96.4
98.1
99.5
F9
F10
F11
F12
1.7
1.4
a
Calculated on Systat 11.
3.2. Interpretation of MODEL 2
jects that configure the final equation. This procedure is known as
CV or internal validation by leave-one-out (LOO). In recent years,
the LOO press statistics have been used as a means of indicating
predictive ability, and many authors consider high values for these
statistics as an indicator, or even as the ultimate proof, of the high
predictive power of a QSAR model.
On the other hand, MODEL 2 shows that the Activated partial
thromboplastin time increases with the increase in the Heat of For-
mation, SAAverage and Quntaum mechanic dipole along X-axis. The
positive correlation with SAAAverage indicates that the hydropho-
bic interactions are involved in the inhibition of thrombus forma-
tion. Involvement of electronic interactions contributed by
Quntaum mechanic dipole along X-axis are also equivalently impor-
tant for the inhibition of Clotting. Similar to the MODEL 1 H-acceptor
count is negatively contributing to the activity. In addition, the x
component of Dipole and Quntaum mechanic dipole along Y-axis
also showed a negative contribution to anticoagulant activity by
APTT.
3. Discussion
3.1. Interpretation of MODEL 1
The model for anticoagulant activity by estimation of prothrom-
bin time [MODEL 1], using Principal Component Regression
Figure 2. A dendrogram illustrating the results for the hierarchical k-NNCA of the set of compounds for prothrombin time in the training and prediction sets of the present
work.

1658
M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
Figure 3. A dendrogram illustrating the results for the hierarchical k-NNCA of the set of compounds for activated partial thromboplastin time in the training and prediction
sets of the present work.
more than 500 descriptors, which delineate liphophilic, conforma-
tional, electronic, spatial, structural, thermodynamic and quantum
mechanical information. For the calculation of quantum descrip-
tors MOPAC simulations were used with the PM3 Hamiltonian.
4. Conclusion
The elongation of prothrombin time due to the addition of the
pyridyl chromon-2-one derivatives indicates that the synthesized
derivatives have anticoagulant properties. The mechanistic data ob-
tained from the elongation of the prothrombin time indicates that
the activity may be due to inhibition of clotting factors II, V, VII
and X. While the elongation in activated partial prothrombin time
may be due to inhibition of clotting factors V, VIII, IX, X and XIII.
The initial and maximum effect of anticoagulants is based on the
half-lives of each of these factors. Factor VII has a half-life of 6 h
so the effect of coumarin on this factor will increase the bleeding
to a certain degree within 6 h, however Factor II and X exhibit
half-lives of 48–72 h, so the maximum effect of coumarin is not
seen until three days after initiation or dose change. The coumarin
types of drugs act by inhibition of factor. Thus inspection of actual
interacting site and the interactions at the active site of the clotting
factors can lead to discovery potent anticoagulant agents.
5.2. Step-I: synthesis of ethyl-2-(6-aminopyridin-3-yl)-3-
oxobutanoate
A mixture of 2 g of potassium hydroxide, 3 g of potassium car-
bonate and 3 mL of ethyl acetoacetate (0.01 mol) was irradiated
under microwave irradiation for 2 min at 455 W. After the forma-
tion of potassium salt of ethyl acetoacetate the solution of 4g
(0.01 mol) 2-amino-5-bromopyridine in 10 mL DMSO was added,
and the mixture was stirred vigorously under microwave irradia-
tion for 40 min at 700 W. The mixture was cooled and the DMSO
was removed under vacuum. The product was extracted in three
25 mL portions of chloroform all the fractions were combined
and chloroform was evaporated to obtain the product which was
recrystallized from methanol.
5. Experimental
5.3. Step-II: synthesis of 3-(6-aminopiridin-3-yl)-4-methyl-2H-
chromen-2-one19–22
5.1. Materials and methods
The IR (KBr) spectra of the synthesized compounds were re-
corded on a Jasco- FTIR 4100 instrument. The 1H NMR spectra of
the compounds were recorded on 400 MHz Varian NMR, using
DMSO-d6 as solvent. This series was subjected to QSAR studies
using Vlife Molecular Design Suite (MDS) 3.0 (obtained from Vlife
sciences, Pune) running on P-IV processor. The quantum and phys-
icochemical properties of each compound were specified using
In 500 mL three-necked round bottom flask, with sealed mag-
netic stirrer, a thermometer reaching to bottom placed 30 mL of
concentrated sulfuric acid. The flask was surrounded with ice bath
and when temperature falls below 10 °C added a mixture of 1 g
(2 mol) of substituted phenol and 3.2 g (2 mol) of ethyl-2-(6-
aminopyridin-3-yl)-3-oxobutanoate the mixture was stirred, and
the temperature kept below 10 °C by means of ice and salt. After

M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
1659
Table 4
Table 5
Table showing the results of the observed and predicted biological activity for
prothrombin time (PT)
Table showing the results of the observed and predicted biological activity for
activated partial thromboplastin time (APTT)
Sr. no
Observed PIC50 PT
Predicted PIC50 PTa
Residualsa
Sr. No
Observed PIC50 APTT
Predicted PIC50 APTTa
Residualsa
1
2b
3
4
5
6
7
8
9
0.461
0.028
0.250
0.244
0.473
0.455
0.326
0.161
0.857
0.782
0.792
0.657
0.392
0.666
0.590
0.384
0.554
0.311
0.447
0.364
0.783
0.526
0.559
0.785
0.812
0.682
0.594
0.751
0.350
0.459
0.808
0.210
0.326
0.442
0.759
0.252
0.610
0.774
0.658
0.791
0.615
0.780
0.493
0.494
0.724
0.366
0.398
0.310
0.612
0.309
0.573
0.684
0.461
0.693
0.407
0.831
0.402
0.350
0.656
0.182
0.211
0.226
0.705
0.290
0.448
0.126
0.263
0.072
0.498
0.385
0.221
0.031
0.709
0.690
0.728
0.728
0.476
0.638
0.516
0.436
0.538
0.368
0.493
0.390
0.742
0.503
0.563
0.771
0.632
0.684
0.526
0.800
0.436
0.423
0.573
0.393
0.497
0.373
0.590
0.197
0.680
0.802
0.706
0.791
0.641
0.751
0.561
0.330
0.736
0.372
0.473
0.360
0.566
0.283
0.563
0.671
0.574
0.629
0.408
0.900
0.563
0.377
0.602
0.327
0.365
0.139
0.585
0.232
0.012
ꢀ0.098
ꢀ0.013
0.171
1
2b
3
4
5
6
7
8
9
0.439
ꢀ0.098
0.076
0.070
0.392
0.154
0.186
ꢀ0.060
0.532
0.537
0.489
0.433
0.291
0.521
0.289
0.344
0.403
0.163
0.493
0.370
0.304
0.025
0.280
0.434
0.446
0.347
0.311
0.401
0.049
0.158
0.507
0.190
0.202
0.141
0.318
ꢀ0.049
0.275
0.533
0.385
0.490
0.396
0.400
0.239
0.193
0.523
0.265
0.097
0.159
0.483
0.099
0.158
0.263
0.308
0.292
0.306
0.380
0.101
0.249
0.455
0.051
0.110
0.486
0.054
0.073
0.348
ꢀ0.026
0.138
ꢀ0.099
0.287
0.190
0.085
ꢀ0.083
0.430
0.448
0.471
0.541
0.356
0.473
0.247
0.308
0.380
0.186
0.534
0.402
0.463
0.171
0.293
0.423
0.301
0.396
0.305
0.293
0.102
0.153
0.449
0.197
0.236
0.221
0.291
ꢀ0.076
0.364
0.506
0.419
0.487
0.368
0.332
0.229
0.084
0.392
0.281
0.194
0.249
0.431
0.160
0.100
0.301
0.217
0.338
0.231
0.329
0.100
0.320
0.485
0.121
0.176
0.334
0.040
0.129
0.090
ꢀ0.071
ꢀ0.062
0.169
ꢀ0.025
0.104
0.069
ꢀ0.037
0.104
0.129
0.147
0.091
0.101
0.023
0.101
0.088
0.017
ꢀ0.108
ꢀ0.065
0.047
0.041
0.035
10
11
12
13
14
15
16b
17
18
19
20b
21
22
23b
24
25b
26
27
28
29
30
31
32
33
34
35b
36
37
38
39
40
41
42
43
44
45
46
47
48b
49
50
51
52b
53
54
55
56
57b
58b
59
60
61b
62
63b
64
10
11
12
13
14
15
16b
17
18
19
20b
21
22
23b
24
25b
26
27
28
29
30
31
32
33
34
35b
36
37
38
39
40
41
42
43
44
45
46
47
48b
49
50
51
52b
53
54
55
56
57b
58b
59
60
61b
62
63b
64
0.063
ꢀ0.071
ꢀ0.084
0.027
0.073
ꢀ0.052
0.015
0.022
ꢀ0.057
ꢀ0.046
ꢀ0.026
0.040
ꢀ0.023
ꢀ0.041
ꢀ0.032
ꢀ0.159
ꢀ0.146
ꢀ0.013
0.010
0.022
ꢀ0.004
0.013
0.179
0.144
ꢀ0.002
ꢀ0.0496
0.067
0.005
ꢀ0.049
ꢀ0.086
0.035
0.107
ꢀ0.053
0.004
0.234
0.057
ꢀ0.183
ꢀ0.171
0.068
0.168
0.054
ꢀ0.070
ꢀ0.028
ꢀ0.048
0.000
ꢀ0.007
ꢀ0.034
ꢀ0.080
0.026
0.027
ꢀ0.089
0.026
ꢀ0.034
0.002
0.027
0.067
0.009
0.108
0.130
ꢀ0.016
ꢀ0.097
ꢀ0.090
0.051
ꢀ0.026
0.028
ꢀ0.068
0.163
ꢀ0.012
ꢀ0.006
ꢀ0.075
ꢀ0.050
0.045
0.025
0.009
0.012
ꢀ0.061
0.057
ꢀ0.039
0.090
ꢀ0.113
0.063
ꢀ0.046
ꢀ0.001
ꢀ0.069
ꢀ0.161
ꢀ0.027
0.053
ꢀ0.145
ꢀ0.154
0.086
0.074
0.050
0.001
ꢀ0.071
ꢀ0.030
ꢀ0.070
ꢀ0.066
0.151
0.013
ꢀ0.056
0.119
0.057
a
a
Calculated on Vlife Sciences Molecular design Suite 3.0.
Indicates molecules of test set.
Calculated on Vlife Sciences Molecular design Suite 3.0.
Indicates molecules of test set.
b
b
all the mixture had been added (about 2 h) the reaction mixture
was stirred for 24 h without further cooling. The reaction mixture
was poured with vigorous stirring into ice and water mixture. The
precipitate was then collected on filter paper and washed with
50 mL portions of cold water. The crude product was dissolved in
50 mL of 5% NaOH solution and the solution was filtered, and
substituted coumarin was reprecipited from the filtrate by slow
addition of dilute (1:10) sulfuric acid until solution until the solu-
tion was neutral to litmus. During neutralization the mixture must
be well stirred. The product was collected on Buchner funnel and
washed with four 50 mL portions of cold water and dried. IR
(KBr), 1725–1720(C@O); 1150–1200(C–O–C); 1360–1310(C–N);

1660
M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
o Pipette 0.1 mL of plasma in small test tube; add 0.1 mL of
Correlation Plot for Prothrombin Time
(PT)
100 mg/mL test compound solution. Incubate for 5 min and
repeat the procedure to record the elongation in mean pro-
thrombin time.
1
0.8
0.6
0.4
0.2
0
o The IC50 values were calculated from the dose response curve.
5.6. Elongation of activated partial thromboplastin time
(APTT)27
o Pipette out 0.1 mL of brain extract, 0.1 mL of kaolin reagent and
0.2 mL of plasma in a test tube. Incubate for 1 min at 37 °C.
o Add 0.1 mL of calcium chloride and start the stop watch.
o After 20 s observe the formation of clot by tilting the test tube.
As soon as the clot is observed note the time as control reading.
o Pipette out 0.1 mL of brain extract, 0.1 mL of kaolin reagent and
0.2 mL of plasma in small test tube; add 0.1 mL of 100 mg/mL
test compound solution. Incubate for 5 min and repeat the pro-
cedure to record the elongation in mean activate partial throm-
boplastin time.
0
0.2
0.4
0.6
0.8
1
Observed Activity
Training Set
Test Set
Figure 4. Figure showing correlation plot between observed and predicted activity
for prothrombin time (PT).
Correlation Plot For APTT
o The IC50 values were calculated from the dose response curve.
0.6
0.4
0.2
0
5.7. QSAR studies
5.7.1. Data Set
To obtain mathematical expressions able to discriminate be-
tween good (high) and moderate-low activity compounds, the
chemical information contained in a great number of compounds,
with and without the desired biological profile, must be statisti-
cally processed. In the present study, the anti-coagulant activity
was used to establish a classical linear-classification-based QSAR
equation. Taking into account that the most critical aspect, in the
construction of a training dataset, is the molecular diversity of
the included compounds, we had selected compounds having as
much structural variability as possible. The anti-coagulant activity
was evaluated on basis of elongation of prothrombin time and
thrombin time of each compound in the database was evaluated
in vitro. The builder module of the Vlife MDS 3.0 program (Vlife
Sciences Technologies Pvt. Ltd) was used to generate molecular
models of 64, 3-[2-amino(pyridine-3-yl)]chromen-2-one deriva-
tives. These molecular models were built using the template of
the most active molecule from the data set, that is, 3(6-amino-5-
nitro pyridine-3-yl) 4-methyl-2H-chromen-2-one. They were then
energy-minimized using the Merck Molecular Force Field (MMFF)
until the root mean square (rms) gradient reached value
0.001 kcal molꢀ1 Å. The chemical information contained in a great
number of compounds, with and without the desired biological
profile, must be statistically processed. In the present study, the
IC50 was used to establish a classical linear-classification based
QSAR equation.
-0.2
0
0.2
0.4
0.6
-0.2
Observed Activity
Training Set Test Set
Figure 5. Figure showing correlation plot between observed and predicted activity
for activated partial thromboplastin time (APTT).
3398–3381(N–H); 1675–1610(C@N); NMR (DMSO) d : 0.94 (s, 3H,
CH3), 2.48 (s, 2H, NH2), 6.91 (s, 1H, 3-pyridine), 7.14, (s, 1H, 4-pyr-
idine), 7.35–8.25 (m, H, aromatic), 8.91 (s, 1H, 6-pyridine).
5.4. Step-III: synthesis of 4-methy-3-(6-[phenyl methylene]
amino pyridine-3-yl)-2H chromen-2-one23–25
In flask containing 15 mL of glacial acetic acid 1.33 g (0.005 mol)
of 3-(6-aminopiridin-3-yl)-4-methyl-2H-chromen-2-one was added,
to this mixture 0.53 g (0.005 mol) of aldehyde was added. And the
reaction mixture was treated with microwave irradiation of
10 min at 700 W. The mixture after cooling was poured in cold
water the imines precipitate out. IR (KBr), 1725–1720(C@O);
1150–1200(C–O–C); 1360–1310(C–N); 1675–1610(C@N); NMR
(DMSO) d : 0.92 (s, 3H, CH3), 6.99 (s, 1H, 3-pyridine), 7.15, (s, 1H,
4-pyridine), 7.40–8.25 (m, H, aromatic), 8.88 (s, 1H, 6-pyridine),
9.20 (s, 1H, CH@N).
5.7.2. Molecular descriptors for QSAR analysis
A large number of MDs is usually used in QSAR methods. The
specific biological action of drugs is frequently described by hydro-
phobic, electronic, and steric properties. The hydrophobic proper-
ties express the ability of a molecule to be transported through
the organism to interact with biological membranes and to be
bound to the receptor by van der Waals forces. We considered as
hydrophobic descriptor the decimal logarithm of the octanol–
water partition coefficient (slogP). Both electronic and steric prop-
erties characterize the pharmacodynamic properties in the ligand–
receptor interaction. They define the ability of the drug to join to
the receptor. The electronic descriptors calculated by quantum
mechanical procedures for the 2D QSAR Studies were: molecular
weight (molwt), Volume (Vol), H-bond acceptor count (H-acc), H-
5.5. Determination of prothrombin time (PT) by Quick‘s
method26
o Pipette 0.1 mL of plasma in small test tube. Add 0.1 mL of brain
thromboplastin and mix.
o Wait for 2 min and add pre warmed calcium chloride solution
at 37 °C, mix and start the stop watch.
o Hold the tube in front a source of light and keep tilting the test
tube gently. At first appearance of fibrin clot stop the watch
immediately and record the time. Record this time as control
reading.

M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
1661
bond doner count (H-don), Chi3cluster (c3c), chiV3Cluster (cV3c),
dipole moment (DM), QMDipoleX (QMDx), QMDipoleY (QMDy),
QMDipoleZ (QMDz), Information based descriptor (Idaverage),
Quadrupole1 (Q1), Quadrupole2 (Q2), Heat of Formation (HF),
highest occupied molecular orbital (HOMO), lowest unoccupied
molecular orbital (LUMO), ionization potential (IP), molar refrac-
tivity (SMR), partition coefficient (slogP), Moment of inertia along
x-axis (MIx) Moment of inertia along y axis (MIy), Moment of iner-
tia along z-axis (MIz) average hydrophobicity by kellog method
(SKaverave), average hydrophobicity by Audry method (SAaver-
age), Xcomponent of Dipole (Xdipole), Ycomponent of Dipole (Ydi-
pole), Zcomponent of Dipole (Zdipole), radius of gyration (RGYR),
Distanc Topological (DistTop), Kappa1 (K1), Kappa2 (K2), quantum
mechanics dipole moment (QMDM).
and these methods are implemented in either an agglomerative
(bottom–up) or divisive (top–down) procedure. On the other hand,
the partitional clustering assumes that the objects have nonhierar-
chical characters.12–16 Most popular partitional cluster algorithms
are k-mean cluster algorithms (k-MCAs) and another by Jarvis
and Patrick (also known as k-nearest neighbor cluster algorithm;
k-NNCA) algorithms. The k-mean clustering algorithms use an
interchange (or switching) method to divide n data points into k
groups (clusters) so that the sum of distances/dissimilarities
between the objects within the same cluster is minimized. The
k-mean approach requires that k (the number of clusters) is known
before clustering. The Jarvis–Patrick method requires that the user
specify the number of nearest neighbors, as well as the number of
neighbors in common to merge two objects. The Jarvis–Patrick
method is a deterministic algorithm; it does not require iterations
for computations.12–16 To design the training and test series, as well
as to demonstrate the structural diversity of the present database,
we carried out one of these kinds of cluster analyses (k-NNCA) for
anti-coagulant series. The Vlife graph package from the QSAR Mod-
ule was used to develop these CA‘s. In this study, we used the ‘aver-
age linkage’ metric as the method to merge objects into clusters.
The average linkage distance between two clusters is defined as
the average (Euclidean squared arithmetic mean) distance between
pairs of objects, one in each cluster. Average linkage tends to join
those clusters with small variances and produces new clusters with
roughly the same variance.
5.7.3. Analysis of principal components
To conduct the comparison of the MDs computed in this work,
we performed a factorial analysis, using the principal components
method. The theoretical aspects of this statistical technique have
been extensively exposed in the literature including many chemi-
cal applications.5–11 The main uses of factorial analytical tech-
niques are: (1) to reduce the number of variables, and (2) to
detect structure in the relationships between variables, namely,
to classify variables10,11 In this approach, factorial loadings (or
‘new’ variables) are obtained from original variables of Molecular
Descriptors. Thus, these factors capture all the ‘essence’ of these
MDs, because they are linear combinations of the original items.
Because each consecutive factor is defined to maximize that vari-
ability not captured by the preceding factor, consecutive factors
are independent of each other. Put in another way, consecutive fac-
tors are uncorrelated or orthogonal to each other. The first ob-
tained factor is generally more highly correlated with the
variables than the other factors. This is to be expected, because
these factors are successively extracted and will account for less
and less overall variance. The factor analysis was carried out using
‘varimax normalized’ as rotational strategy to obtain the factorial
loadings from the principal component analysis. The goal of this
rotational procedure is to obtain a clearer pattern of the loadings,
that is, factors that be somehow clearly marked by high loadings
for some variables and low loadings for others. The ‘varimax nor-
malized rotation’ is the method that is most commonly used as
‘varimax’ rotation.18 This rotational strategy is aimed at maximiz-
ing the variances of the squared normalized factorial loadings (row
factorial loadings divided by squared roots of the respective com-
munalities), across the variables for each factor. This strategy
makes the structure of the factorial pattern as simple as possible,
permitting a clearer interpretation of the factors without loss of
orthogonality between them. Finally, some of the most important
conclusions, which could can be drawn from a factor analysis that
will be of large usefulness in the present article are the following
(1) variables with a high loading in the same factor are interrelated
and will be the more, the higher the loadings, (2) no correlation ex-
ists between variables having nonzero loadings only in different
factors. These are the principal ideas that permit the interpretation
of the factorial structure, obtained using the factorial analysis as a
classification method, and (3) only variables with high loadings in
different factors may be combined in a regression equation to elim-
inate collinearities.
5.7.5. LDA and classification-based QSAR model
The discriminant functions were obtained using LDA,17 as
implemented in Vlife model building Wizard, the default parame-
ters of this program were used in the development of the model.
Forward stepwise was fixed as the strategy for variable selection.
The principle of maximal parsimony (Occam’s razor) was taken
into account as the strategy for model selection. In its original
form, Occam’s razor states that ‘Entities should not be multiplied
beyond necessity’. In this case, simplicity is loosely equated with
the number of parameters in the model. If we understand the pre-
dictive mistake to be the error rate for unseen examples, the Oc-
cam’s razor can be stated for the selection of QSAR models as
(‘QSAR Occam’s Razor’): Given two QSAR models with the same
predictive error, the simplest one should be preferred because
simplicity is desirable in itself.28 Relation to this, we select that
model with the highest statistical signification, but having as
few parameters as possible.
5.7.6. Validation of the obtained model
The statistical robustness and predictive power of the obtained
model were assessed using a prediction (test) set.29 In addition; a
leave-group-out (LGO) cross-validation (CV) strategy was carried
out. Eliminate a compound in the training set and predict its bio-
logical activity on the basis of the k-NN principle, that is, as the
weighted average activity of k most similar molecules. The simi-
larities are evaluated as Euclidean distances between compounds
using only the subset of descriptors that corresponds to the cur-
rent model. Repeat until every compound in the training set has
been eliminated and its activity predicted once. In this way,
every observation was predicted once (in its group of left-out
observations).
5.7.4. Cluster analysis
Cluster analysis (CA) encompasses a number of different classi-
fication algorithms, and it permits to organize the observed data
into meaningful structures. Many CA algorithms have been in-
vented, and they belong to two categories: hierarchical clustering
and partitional (nonhierarchical) clustering. Hierarchical clustering
rearranges objects in a binary tree-structure (joining clustering),
Acknowledgement
The authors are thankful to Dr. H.N. More, Principal, Bharati
Vidyapeeth College of Pharmacy, Kolhapur for providing facilities
to carry out the research work.

1662
M. S. Bhatia et al. / Bioorg. Med. Chem. 17 (2009) 1654–1662
16. Montero-Torres, A.; Celeste Vega, M.; Marrero-Ponce, Y.; Rolo´n, M.; Go´mez-
References and notes
´
´
´
´
Barrio, A.; Escario, J. A.; Aran, V. J.; Martynez-Fernandez, R.; Meneses-Marcel, A.
Bioorg. Med. Chem. 2005, 13, 6264.
1. Coleman, R. W.; Marder, V. J.; Salzman, E. W.; Hirsh, J.. In Coleman, R. W.,
Marder, V. J., Salzman, E. W., Hirsh, J., Eds., 3rd ed.; J.B. Lippincott: Philadelphin,
1994; pp 3–18.
17. Van de Waterbeemd, H. In Chemometric Methods in Molecular Design; van de
Waterbeemd, H., Ed.; VCH Publishers: Weinheim, Germany, 1995; pp 265–
288.
18. STATISTICA, Statsoft, Inc. 1999. http://www.statsoft.com.
19. Horning, E. C.. In Organic Synthesis Collective; John Wiley & sons, Inc.: USA,
1955; Vol. III, pp 281–283.
20. Rajendra, E.; Ramesh, P.; Ramu, K. Ind. J. Chem. 2004, 43B, 2650.
21. Wang, L.; Xia, J.; Ian, H.; Qiun, C.; Yun, M. S. Ind. J. Chem. 2003, 42B, 2097.
22. Jin, T. S.; Gua, T.; Shauangli, T.; Yin, Y.; Liu, H. Ind. J. Chem. 2003, 42B,
2612.
23. Delia, T. J.; Wilcox, T. M.; Ottoman, R. R. J. Heterocycl. Chem. 1979, 16,
1647.
2. Elodi, S.; Varadi, K. Thromb. Res. 1979, 15, 617.
3. Sturzebecher, J.; Markwardt, F.; Walsmann, P. Thromb. Res. 1976, 9, 637.
4. Rameshwar, N.; Krishna, K.; Ashok Kumar, B.; Parthasarathy, T. Bioorg. Med.
Chem. 2006, 14, 319.
5. Marrero Ponce, Y. J. Chem. Inf. Comput. Sci. 2004, 44, 2010.
6. Marrero-Ponce, Y. Bioorg. Med. Chem. 2004, 12, 6351.
7. Cramer, R. D., III J. Am. Chem. Soc. 1980, 102, 1849.
8. Cramer, R. D., III J. Am. Chem. Soc. 1980, 102, 1837.
9. Needham, D. E.; Wei, I.-C.; Seybold, P. G. J. Am. Chem. Soc. 1988, 110, 4186.
10. Malinowski, E. R.; Howery, D. G. In Factor Analysis in Chemistry; Wiley:
Interscience, New York, 1980.
11. Estrada, E.; Rodryguez, L. J. Chem. Inf. Comput. Sci. 1999, 39, 1037.
12. Xu, J.; Hagler, A. Molecules 2002, 7, 566.
13. Mc Farland, J. W.; Gans, D. J. In Chemometric Methods in Molecular Design;
Waterbeemd, H., Ed.; VCH Publishers: Weinheim, Germany, 1995; pp 295–307.
14. Johnson, R. A.; Wichern, D. W. In Applied Multivariate Statistical Analysis;
Prentice-Hall: Upper Saddle River, NJ, 1988.
24. Morrison, R. T.; Boyd, R. E. In Organic Chemistry, 6th ed.; New Delhi, Prentice-
Hall of India Private Ltd: New Delhi, 1996; pp 657–675.
25. Loudon, G. M. In Organic Chemistry, 4th ed.; Oxford University press: New York,
2002; pp 874–880.
´
´
26. Godkar, P. B.; Godkar, P. B. In Textbook of Medical Laboratory Technology, 2nd
ed.; Bhalani publishing house: Mumbai, India, 2005; pp 840–841.
27. Godkar, P. B.; Godkar, P. B. In Textbook of Medical Laboratory Technology, 2nd
ed.; Bhalani publishing house: Mumbai, India, 2005; p 843.
28. Estrada, E.; Patlewicz, G. Croat. Chim. Acta 2004, 77, 203.
29. Wold, S.; Erikson, L. In Chemometric Methods in Molecular Design; van
15. Marrero-Ponce, Y.; Meneses-Marcel, A.; Machado-Tugores, Y.; Montero Pereira,
´
´
´
´
D.; Escario, J. A.; Nogal-Ruiz, J. J.; Ochoa, C.; Aran, V. J.; Martynez-Fernandez, A.
R.; Garcy´a Sa´nchez, R. N.; Montero-Torres, A.; Torrens, F. Curr. Drug Disc.
´
de Waterbeemd, H., Ed.; VCH Publishers: Weinheim, Germany, 1995;
309.
p
Technol. 2005, 2, 245.