A Journal of Software Engineering and Applications, 2012, 5, 84-87
doi:10.4236/jsea.2012.512b017 Published Online December 2012 (http://www.scirp.org/journal/jsea)
Pharmacophore Model Generation of thrombin Inhibitors
Xing Wang1, Yanling Zhang1, Yuhong Xiang2, Zhenzhen Ren1, Yanjiang Qiao1*
1School of Chinese Materia Medica, Beijing University of Chinese Medicine, Beijing 100102, China; 2Department of
Chemistry, Capital Normal University, Beijing 100048, China.
Received 2012
ABSTRACT
Thrombin, an important factor of clotting system, take part in a variety of physiological actions, such as blood clotting,
anticoagulation, thrombosis and fibrinolysis. Inhibiting thrombin is a pivotal and effective step for the prophylaxis of
venous and arterial thrombosis, as well as prevent myocardial infarction for high-risk patients. In this study, a three di-
mensional pharmacophore model was generated for the molecules which are responsible for vasodilation activities tar-
geting thrombin. Ten compounds with known thrombin-inhibiting activity values were selected as training set to gener-
ate the hypothesis using GALAHAD program in SYBYL 7.0 software. The best hypothesis comprises five pharma-
cophore features: four hydrophobes groups and one positively charged group. It has been further validated towards a
test set including known activity compounds obtained from Binding Database, the values of effectively active hit A%
and comprehensive evaluation index CAI are respectively 63.33% and 2.34, the pharmacophore was proven to be suc-
cessful in discriminating active and inactive inhibitors. Furthermore, the pharmacophore model was used as a 3D
query to screen TCMD (Version 2009) database and 6 hit compounds of higher predicted activity were the reported
cardiovascular activities, which may be useful for further study.
Keywords: Cardiovascular Disease; Thrombin; Pharmacophore; Virtual Screening
1. Introduction
Thrombin, a "trypsin-like" serine protease protein, is an
important factor of clotting system. It take part in a vari-
ety of physiological actions, such as blood clotting, anti-
coagulation, thrombosis and fibrinolysis [1-3]. Moreover,
previous studies have demonstrated that thrombin is one
of the key factors for cancer developing, such as PAR1,
an important type of thrombin, mediates the enhance-
ment effects of thrombin in angiogenesis, invasion and
metastasis, growth of cancer [4-6]. These facts have
raised great interests to find an effective, safe, and orally
available thrombin inhibitors, such as hirudin [7], dabi-
gatran [8], rivaroxaban and apixaban [9], which could be
useful anticoagulant drugs for the prophylaxis of venous
and arterial thrombosis, as well as prevent myocardial
infarction for high-risk patients [10].
In this paper, we have developed a pharmacophore
model whose purpose is to identify the critical pharma-
cophoric features necessary for potent thrombin inhibi-
tors. Further, hypothesis was evaluated by a test set with
known activity compounds, the effectively active hit A%
and comprehensive evaluation index CAI were used to
assess the model. Finally, the pharmacophore model was
used as a 3D query to screen TCMD (Version 2009) da-
tabase, 52 compounds were hit and the 6 hit compounds
of higher predicted activity were the reported cardiovas-
cular activities, which may be useful for further study.
2. Materials and Methods
2.1. Compounds and biological data
Compounds 1～30, which can inhibit thrombin, were
taken from the literatures [11-13] and served as the data-
base in the pharmacophore modeling. The structures and
inhibitory activities are listed in Figure 1. The chemical
structures were drawn in ISIS-Draw software and saved
in SYBYL mol2 format, then all the 2D structures were
converted to 3D structures in SYBYL 7.0 software.
N
HN
O
H
N
SO
O
Compound 1 Compound 2 Compound 3

Pharmacophore Model Generation of thrombin Inhibitor
85
(Ki 23nM) (Ki 2nM) (Ki 7nM)
HN
H
N
N
N
NH
O
H2N
NH
H2N
O
N
H
Compound 4
(Ki 170nM)
Compound 5
(Ki 5640nM)
Compound 6
(Ki 2139nM)
NH
N
NNH
NH2
Compound 7
(Ki 191nM)
Compound 8
(Ki 503nM)
Compound 9
(Ki 2655nM)
N
NH
NH2
N
NH
NH2
Compound 10
(Ki 1136nM)
Compound 11
(Ki 1099nM)
Compound 12
(Ki 2124nM)
N
NH
NH2
Compound 13
(Ki 996nM)
Compound 14
(Ki 48nM)
Compound 15
(Ki 1790nM)
Compound 16
(Ki 5269nM)
Compound 17
(Ki 1796nM)
Compound 18
(Ki 3700nM)
N
H
NO
H
N
NH
OO O
O
NH2
O
N
O
H
N
H
N
NH
OO O
O
NH2
O
Compound 19
(Ki 21300nM)
Compound 20
(Ki 20600nM)
Compound 21
(Ki 12100nM)
NH
O
NH2
Compound 22
(Ki 132nM)
Compound 23
(Ki 237nM)
Compound 24
(Ki 920nM)
NH
N
NNH
NH2N
H
NO
H
N
NH
OO O
O
O
NH2
O
O
Compound 25
(Ki 395nM)
Compound 26
(Ki 23300nM)
Compound 27
(Ki 21600nM)
Compound 28 Compound 29 Compound 30
(Ki 1890nM) (Ki 14900nM) (Ki 16300nM)
Figure 1. Chemical structure of thrombin inhibitors.
2.2. Modeling tool
All the pharmacophore modeling calculations were car-
ried out by using the GALAHAD module implemented
in SYBYL 7.0 (Tripos Inc.) software, the UNITY mod-
ule was used to perform a flex search for the potential
inhibitors. All the tasks were running on Xeon Intel(R)
X5460 CPU 3.16 GHz, RAM Memory 12 GB under the
Windows XP system.
2.3. Pharmacophore studies
GALAHAD (Genetic Algorithm with Linear Assignment
of Hypermolecular Alignment of Datasets) uses Tripos'
proprietary technology [14-15] to generate pharma-
cophore hypotheses and alignments from sets of ligand
molecules that bind at a common target site. Before es-
tablishing the model, all the structures of ligands were
checked, hydrogen atoms were added and a minimization
procedure was implemented using the MMFF94
force-field. GALAHAD was run for 130 generations
with a population size of 100. The rest of the parameters
were set as default values. Only models with 8 ligands
with contribution to the consensus feature were consid-
ered. The generated models were evaluated by a test da-
tabase containing 111 compounds, including 30 active
compounds and 81 non-active compounds.
2.4. Virtual Screening
The pharmacophore models built within the Catalyst
software can be used as queries for 3D database screen-
ing. Virtual screening of such databases can serve two
main purposes: first, validating the quality of the gener-
ated pharmacophore models by selective detection of
compounds with known thrombin inhibitory activity, and
second, finding novel, potential leads suitable for further
development. In this study, the best pharmacophore
model was used to search TCMD (Version 2009) to quest
for the potential inhibitors of thrombin from traditional
Chinese medicine.
3. Results and Discussion
3.1. Generation of pharmacophore hypotheses
GALAHAD models were derived by using ten active
ligands as a training set [compounds 4-6, 8, 9, 15, 17, 18,
25, 28], 20 pharmacophore models were derived by us-
ing the training set after GALAHAD run. All the 20
models were evaluated successively by the test database
constructed previously. Table 3 shows the predictable
results of the test database for models with all the eleven
ligands with contribution to the consensus feature.

Pharmacophore Model Generation of thrombin Inhibitor
86
In order to intuitively understand the meaning of indi-
cators used to evaluate the performance of the models,
the schematic diagram was listed (Figure 2) and the pa-
rameter values for each pharmacophore model generated
were listed in Table 1. D is for the total number of com-
pounds in test database and A represents the number of
active compounds. Ht is the total number of hit com-
pounds from test database and Ha represents the number
of active hit compounds from test database, A% repre-
sents the ability to identify active compounds from test
database, Y% represents the proportion of active com-
pounds in the hit compounds. N, the index of effective
identification, is used to evaluate the ability of the mod-
els to identify active compounds from the non-active
compounds. CAI, a comprehensive evaluation index, is
used to identify the best pharmacophore model. Higher
value of CAI is considered to be the better model. In this
study, MODEL_016, with the highest value of CAI was
considered to be the best model.
Figure 2. The schematic diagram of indicators used to
evaluate the pharmacophore mode l
Table 1. The parameter values for each pharmacophore model
Ht Ha A% Y% N CAI
MODEL_01 12 12 40.00 100.00 3.70 1.48
MODEL_02 17 17 56.67 100.00 3.70 2.10
MODEL_03 14 14 46.67 100.00 3.70 1.73
MODEL_04 19 19 63.33 100.00 3.70 2.34
MODEL_05 9 9 30.00 100.00 3.70 1.11
MODEL_06 29 21 70.00 72.41 2.68 1.88
MODEL_07 17 17 56.67 100.00 3.70 2.10
MODEL_08 16 16 53.33 100.00 3.70 1.97
MODEL_09 6 6 20.00 100.00 3.70 0.74
MODEL_10 19 19 63.33 100.00 3.70 2.34
MODEL_11 17 17 56.67 100.00 3.70 2.10
MODEL_12 19 19 63.33 100.00 3.70 2.34
MODEL_13 19 19 63.33 100.00 3.70 2.34
MODEL_14 19 19 63.33 100.00 3.70 2.34
MODEL_15141446.67 100.00 3.70 1.73
MODEL_16191963.33 100.00 3.70 2.34
MODEL_17111136.67 100.00 3.70 1.36
MODEL_18171756.67 100.00 3.70 2.10
MODEL_19141446.67 100.00 3.70 1.73
MODEL_20131343.33 100.00 3.70 1.60
MODEL_016 is displayed in Figure 3, cyan and red
spheres indicate hydrophobes group and positively
charged charged group, respectively. MODEL_016 in-
cludes five pharmacophore features: four hydrophobes
groups and one positively charged group.
Figure 3. Pharmacophore MODEL_016 and molecular
alignment of the compounds used to elaborate the model.
3.2. Virtual Screening
Pharmacophore MODEL_016 was used to screen TCMD
(Version 2009). TCMD screening yielded a hit list of 52
compounds and the 6 hit compounds of higher predicted
activity (Bufotoxin, Cyclovirobuxine, Cyclovirobuxine
D, Leurosine, Methoxyadiantifoline and Thaliadanine)
were the reported cardiovascular activities, such as vaso-
dilator, anti-arrhythmic, cardiac, coronary vasodilator
and anti-hypertension, et al. Cyclovirobuxine mapped to
MODEL_016 is showed in Fig 4.
Figure 4. MODEL_016 mapped with Cyclovirobuxine
4. Conclusions and Future Work
In this study, we have built the pharmacophore model of
thrombin inhibitors using GALAHAD module imple-
mented in SYBYL 7.0 software. The optimal pharma-
cophore model contains 5 pharmacophore features: four
D
A Ha Ht
A%=(Ha/A)×100
Y%=(Ha/Ht)×100
N= (Ha/Ht) / (A/D)
CAI=N×A%

Pharmacophore Model Generation of thrombin Inhibitor
87
hydrophobes group and one positively charged charged
group. The external validation shows that the values of
effectively active hit A% and comprehensive evaluation
index CAI are respectively 63.33% and 2.34, which can
prove that the pharmacophore model is reliable and
available. The virtual screen results of TCMD (Version
2009) database shows that the pharmacophore model has
the ability of determining vasoactive compound related
to thrombin, and the validation experiments needs to be
further studied.
5. Acknowledgements
This work is financially supported by the Foundation of
National Natural Science Foundation of China (No.
81173522) and National Key Technology R&D Program
(No. 2008BAI51B01) in Beijing University of Chinese
Medicine.
REFERENCES
[1] Sambrano GR, Weiss EJ, Zheng YW, et al. “Role of
thrombin signalling in platelets in haemostasis and
thrombosis”. Nature, Vol. 413, No. 6, 2011, pp. 74-78.
doi:10.1038/35092573
[2] H LUO, Y LIANG. “Thrombin-activatable fibrinolysis
inhibitor: a new modulator linking coagulation and fibri-
nolysis”. Basic Medical Sciences and Clinics, Vol. 23,
No. 5, 2003, pp. 484-489. doi:6325(2003)05-0484-06
[3] Weitz JI. “Factor Xa and thrombin as targets for new oral
anticoagulants”. Thrombosis Research, Vol. 127, No. 2,
2011, pp. S5-S12. doi:10.1016/S0049-3848(10)70147-X
[4] Crawley JT, Zanardelli S, Chion CK, et al. “The
central role of thrombin in hemostasis”. Journal of
Thrombosis and Haemostasis,Vol. 5, Sup. s1, 2007,pp.
95-101. doi:10.1111/j.1538-7836.2007.02500.x
[5] Nierodzik ML,Karpatkin S. “Thrombin induces tumor
growth, metastasis, and angiogenesis: Evidence for a
thrombin-regulated dormant tumor phenotype”. Cancer
Cell, vol.10,No.5,2006, pp.355-362.
doi.org/10.1016/j.ccr.2006.10.002
[6] Bajzar L, Chan AK, Massicotte MP, et al. “Thrombosis
in children with malignancy”. Current Opinion in Pe-
diatrics,vol.18,No.1, 2006 ,pp.1-9.
doi:10.1097/01.mop.0000193270.09001.ea
[7] Eriksson BI, Quinlan DJ, Eikelboom JW. “Novel Oral
Factor Xa and Thrombin Inhibitors in the Management of
Thromboembolism”. Annual Review of Medicine,
vol.62,2011,pp. 41-57.
doi:10.1146/annurev-med-062209-095159
[8] Bendel SD, Bona R, Baker WL. “Dabigatran: an oral
direct thrombin inhibitor for use in atrial fibrillation”.
Advances in Therapy,vol.28,No.6,2011,pp.460-472.
doi:10.1007/s12325-011-0025-1
[9] A. Kobsar , J. Koessler , L. Kehrer,etal. “The thrombin
inhibitors hirudin and Refludan® activate the soluble
guanylyl cyclase and the cGMP pathway in washed hu-
man platelets”. ThrombosisandHaemosta-
sis.Vol.107.No.3,2012,pp.521-529.
doi:10.1160/TH11-07-0461
[10] Steinmetzer T, Sturzebecher J. “Progress in the Devel-
opment of Synthetic Thrombin Inhibitors as New Orally
Active Anticoagulants”.Current Medicinal Chemis-
try,Vol.11,No.17, 2004,pp.2297-2321.
doi:10.2174/0929867043364540
[11] Santosh P, Gregory A. L., Julie E. P. “Conformation
Mining: An Algorithm for Finding Biologically Rele-
vant Conformations”. J. Med. Chem., Vol. 48, No.9, 2005,
pp. 3313-3318. doi:10.1021/jm049066l.
[12] Milan Bruncko, William J. McClellan, Michael D. Wendt,
et al. “Naphthamidine urokinase plasminogen activator
inhibitors with improved pharmacokinetic properties”.
Bioorganic & Medicinal Chemistry Letters, Vol.15,No.1,
2005, pp. 93-98. doi:10.1016/j.bmcl.2004.10.026
[13] Petra Štefanic Anderluh, Marko Anderluh, Janez Ilaš.
“Toward a Novel Class of Antithrombotic Compounds
with Dual Function. Discovery of
1,4-Benzoxazin-3(4H)-one Derivatives Possessing
Thrombin Inhibitory and Fibrinogen Receptor Antagonis-
tic Activities”, J. Med. Chem., 2005, 48 (9), pp
3110–3113. doi:10.1021/jm048984g
[14] Richmond, N.J., Abram, C.A., Wolohan, P.R., et al.
“GALAHAD: 1 Pharmacophore Identification by Hy-
permolecular Alignment of Ligands in 3D”. J COMPUT
AID MOL DES, Vol. 20, No. 9, 2006, pp. 567-587.
doi:10.1007/s10822-006-9082-y.
[15] N.J. Richmond, P.Willett, R.D. Clark. “Alignment of
three-dimensionalmolecules using an image recognition
algorithm”. J.Mol.Graph.Model. Vol. 23, No. 2, 2004, pp.
199-209. doi:10.1016/j.bbr.2011.03.031.