7
4
M.-H. Dong et al. / Bioorg. Med. Chem. 24 (2016) 73–84
venous thromboembolic events, stroke, systemic embolism.17,18
However, its bioavailability is low and has a dose-dependent risk
on windows 7 workstation. Each molecule (creation of training set,
test set and docking calculations) was optimized using Tripos force
7
,19,20
36
37
of haemorrhage,
which highlight the critical need to solve
field and Gasteiger-Huckel charges. 3D structures of all com-
pounds were constructed by using the Sketch Molecule module.
Structural energy minimization was performed using Powell gradi-
ent algorithm with a convergence criterion of 0.005 kcal/(mol * Å)
and a maximum of 10,000 iterations. For the CoMFA analysis, steric
and electrostatic fields were generated and scaled by the CoMFA-
STD method with a default cutoff energy value of 30 kcal/mol so
as to decrease the electrostatic energies and set the domination
of large steric to a minimum, calculated at each lattice intersection
on a regularly spaced grid of 3 Å. The grid pattern was generated
the problem of this side effect, newly dabigatran derivatives are
being developed. Therefore, research efforts should aimed at the
identification of synthetic, orally active, dabigatran derivatives.
In recent years, many papers have reported some progresses in
thrombin or its inhibitors in chemical synthesis. Such as de Candia
et al. synthesis direct thrombin inhibitors bearing 4-(Piperidin-1-
9
b
21
yl)pyridine. Chobanian et al. Improved stability of Proline-
derived direct thrombin inhibitors through hydroxyl to heterocycle
replacement. However, performing these experiments is costly and
time-consuming and may result in producing of toxic side effects.
For saving resources and expedite the development of thrombin
inhibitors, computer aided drug design approaches can be used
to offer a deeper insight into inhibitor’s structure or design effec-
tive inhibitors and employed to obtain the biological activity of
newly designed compounds. One of the well-known computational
methods used effectively in drug design is the three-dimensional
quantitative structure–activity relationship (3D-QSAR) without a
3
8
automatically by the SYBYL/COMFA routine. In the case of CoM-
SIA analysis, similarity indices descriptors were derived with the
same lattice box that was used in CoMFA. In addition to steric
and electrostatic fields, the CoMSIA model result in extra informa-
tion including hydrophobic, H-bond donor and acceptor descrip-
tors. Other parameters were set by default in SYBYL-X 2.0.
2.3. Molecular alignment
2
2–24
doubt.
In this method, the biological activities of compounds
can be predicted based on the molecular properties. The results
Molecular alignment is considered as a crucial step for 3D-QSAR
study. In order to develop a reliable 3D-QSAR model, three differ-
ent alignment rules were adopted. The first rule was ligand-based
alignment (Alignment I). During this process, dabigatran was cho-
sen as template and the remaining compounds were aligned based
on an atom-by-atom fitting principle. Figure 1A and B shows the
atoms of the common substructure used to automatically position
the compounds. The second was the docking-based alignment
(Alignment II). The bioactive conformation of each compound
obtained from molecular docking were aligned automatically and
imported directly into a molecular database for CoMFA and
CoMSIA researches. The result based on alignment II is shown in
Figure 1C. Alignment III was similar to alignment I, but the molec-
ular conformations aligned to dabigatran were gained from molec-
ular docking. The alignment model is described in Figure 1D.
25
of comparative molecular field analysis (CoMFA) and the com-
parative molecular similarity indices analysis (CoMSIA)26 studies
would lead to graphical visualization of key chemical structural
features where the interactive fields (steric, electrostatic,
hydrophobic and hydrogen bond donor/acceptor) may affect the
2
7–30
biological activities.
Dabigatran etexilate is a new type of synthetic DTI that is
3
1,32
expected to play an important role in anticoagulation therapy.
To the best of our knowledge, there has never been data reported
concerning the theoretical calculations study on thrombin inhibi-
tors (Dabigatran derivative). In this work, 3D-QSAR tools like
CoMFA and CoMSIA based on different alignment rules and Topo-
mer CoMFA were constructed to derive the contour maps applied
to determine the structural factors that affect the inhibitory activ-
ity of thrombin inhibitors. In addition to the SAR study, molecular
docking had been carried out to study the possible binding modes
of inhibitors at the active site of thrombin protein. Finally, the
molecular modeling further validated by chemical synthesis and
biological evaluation of two new compounds. The information
derived, we hope, will be of useful in thrombin inhibitors develop-
ment program.
3
3
. Results and discussion
.1. 3D-QSAR analysis
The CoMFA and CoMSIA models were developed based on the
set of 39 compounds, and an test set including 9 compounds were
employed to evaluate the reliability and applicability of the built
model. Partial least squares (PLS) method was used to generate sta-
tistically significant 3D-QSAR models in regression analysis.39 To
assess the predictive ability of the constructed PLS model, cross-
validation analysis was employed using the leave-one-out (LOO)
2
2
. Materials and methods
.1. Data sets
4
0
Molecular structures and biological activities of 48 compounds
method where one of the molecules was excluded from the data
set and its activity was predicted by the model derived from the
involved in this study as thrombin inhibitors were taken from the
33–35
22
literature of our laboratory has reported.
The biological activ-
rest of the data set. The results of cross-validation correlation
2
2
ities as IC50 (nM) values were converted to logarithmic scale pIC50
to give numerically larger data values. The samples were randomly
divided into a training set of 39 compounds (80%) for model gener-
ation and a test set of 9 compounds (20%) for predictability assess-
ment, considering the principle of distribution in the range of the
biological data for the both sets and chemical structures diversity.
The chemical structures and biological activities expressed as pIC50
are given in Table 1. The test set compounds are marked and
shown in Table 1.
coefficient (q ) and the squared correlation coefficient (r ) are good
statistical parameter to show the predictive capability of the model
which were utilized to investigate the robustness and statistical
validity of the built models. In addition, the CoMFA and CoMSIA
2
2
models yielded a q > 0.5 and r > 0.9, indicating that they were
4
1
reliable enough for activity prediction. The results of the built
models using three different alignment rules were summarized
in Table 2. As can be seen, the all q and r values of CoMFA and
CoMSIA models based on alignment I rule were more than those
of obtained from alignment II and III. For CoMFA analysis, with
2
2
2
2
.2. Computational approach
an optimal number of components (ONC) of 12 and q of 0.709.
The non-cross-validated coefficient r , standard error of estimate
2
3
D-QSAR modeling and molecular docking were performed
(SEE) and F-statistic values are 0.992, 0.053 and 273.711, respec-
using the SYBYL-X 2.0 package (Tripos Inc., St. Louis, USA) running
tively. The contributions of the steric and electrostatic fields are