Journal of Molecular Graphics and Modelling 79 (2018) 103–117

Contents lists available at ScienceDirect

Journal of Molecular Graphics and Modelling

journal homepage: www.elsevier.com/locate/JMGM

Topical Perspectives

Structure-based design of hERG-neutral antihypertensive oxazalone

and imidazolone derivatives
Busecan Aksoydan a,1 , Isik Kantarcioglu a,1 , Ismail Erol a,b , Ramin Ekhteiari Salmas a ,
Serdar Durdagi a,∗
a
Computational Biology and Molecular Simulations Laboratory, Department of Biophysics, School of Medicine, Bahcesehir University (BAU), Istanbul,
Turkey
b
Department of Chemistry, Gebze Technical University, Kocaeli, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: Angiotensin II receptor type 1 (AT1) antagonists are the most recent drug class against hypertension.
Available online 18 October 2017 Recently first crystal structure of AT1 receptor is deposited to the protein data bank (PDB ID: 4YAY). In
this work, several molecular screening methods such as molecular docking and de novo design studies
were performed and it is found that oxazolone and imidazolone derivatives reveal similar/better inter-
action energy profiles compared to the FDA approved sartan molecules at the binding site of the AT1
receptor. A database consisting of 3500-fragments were used to enumerate de novo designed imida-
zolone and oxazolone derivatives and hereby more than 50000 novel small molecules were generated.
These derivatives were then used in high throughput virtual screening simulations (Glide/HTVS) to find
potent hit molecules. In addition, virtual screening of around 18 million small drug-like compounds
from ZINC database were screened at the binding pocket of the AT1 receptor via Glide/HTVS method.
Filtered structures were then used in more sophisticated molecular docking simulations protocols (i.e.,
Glide/SP; Glide/XP; Glide/IFD; Glide/QPLD, and GOLD). However, the K+ ion channel/drug interactions
should also be considered in studies implemented in molecular level against their cardiovascular risks.
Thus, selected compounds with high docking scores via all diverse docking algorithms are also screened
at the pore domain regions of human ether-a-go-go-related gene (hERG1) K+ channel to remove the high
affinity hERG1 blocking compounds. High docking scored compounds at the AT1 with low hERG1 affinity
is considered for long molecular dynamics (MD) simulations. Post-processing analysis of MD simulations
assisted for better understanding of molecular mechanism of studied compounds at the binding cavity
of AT1 receptor. Results of this study can be useful for designing of novel and safe AT1 inhibitors.
© 2017 Elsevier Inc. All rights reserved.

1. Introduction IHPF) occurring in the RAS, is associated with increased arterial BP

Hypertension (also known as) high blood pressure (BP) is con-
sidered to be the most prevalent type of cardiovascular risk factors.
BP is associated with stroke, myocardial infarction, heart and
kidney failure. The diagnosis in early stages of hypertension is
impossible due to the rare symptoms and complex pathogene-
sis [1–4]. Renin-angiotensin system (RAS) is a hormone system
responsible for regulation of the BP and fluid balance. Angiotensin II
(A-II), a strong vasoconstrictive octapeptide hormone (seq. DRVY-
DOI of original article: http://dx.doi.org/10.1016/j.jmgm.2017.08.004.
∗ Corresponding author.
E-mail address: serdar.durdagi@med.bau.edu.tr (S. Durdagi).
1
These authors are equally contributed.
[5]. Elucidation of the RAS mechanism, will be able to lead the devel-
opment of novel treatments. Inhibition of this mechanism either
by blocking angiotensin converting enzyme (ACE) or via blocking
angiotensin type II receptor I (AT1) is found to have reducing effect
on BP as well as on the other cardiovascular related diseases [4].
AT1 receptor belongs to the G-Protein Coupled Receptor (GPCR)
family and is formed by seven transmembrane (TM) domains with
intracellular and extracellular loops [5]. Despite many advances in
the treatment of hypertension in recent years, it is reported that
approximately 56% of BP values of patients cannot be controlled,
adequately. In addition, it is still not clear which treatment has to
be applied for the diseases such as diabetes in patients suffering
hypertension. Therefore, it is needed to discover multi-functional
drugs for both hypertension and associated complications [6].

https://doi.org/10.1016/j.jmgm.2017.10.011
1093-3263/© 2017 Elsevier Inc. All rights reserved.
104 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

Scheme 1. Hierarchical screening steps of 18 million small molecules in ZINC Database on AT1 receptor crystal structure (left) and AT1 receptor 3D homology model (right).

AT1 antagonists is the most recent drug class of molecules
against hypertension and they mediate their actions through block-
ing detrimental effects of A-II when acts on AT1 receptor. The effects
of AT1 antagonists are not limited to cardiovascular diseases. It is
reported that AT1 receptor blockers may be used as potential anti-
cancer agents − due to the inhibition of cell proliferation stimulated
by A-II − and also can be used against Alzheimer’s disease [7,8].
Therefore, AT1 receptors and the A-II biosynthesis mechanisms are
targets for the development of new synthetic drugs and therapeutic
treatment of various cardiovascular and such diseases.
In the past few years, several non-peptide AT1 receptor antag-
onists have been described as “sartans”, which are effective in the
treatment of hypertension and cardiovascular disorders [9–15].
These approved drugs differ in their structure, pharmacological
profile and efficacy. There is a great interest in the research field for
the design, discovery and development of novel AT1 antagonists
possessing better pharmacological profile and pharmacokinetic
properties [16,17]. In this work, several molecular screening meth-
ods such as molecular docking and de novo design studies were
performed. The oxazolone and imidazolone derivatives were found
to have similar/better interaction energy profiles compared to FDA
approved sartan molecules at the binding pocket of AT1 receptor.
However, the K+ ion channel/drug interactions should be con-
sidered in studies implemented in molecular level against their
cardiovascular risks. For example, human ether-a go-go related
gene (hERG1) K+ ion channel blockage with different molecules
causes long QT syndrome (LQTS) that induces serious cardiovascu-
lar complications and may even result in sudden death. hERG1K+
ion channel has 3 main conformations (open, open-inactivated, and
closed-states) are modeled by our group in recent years [18–21].
In this study, selected potent ligands were also checked by their
predicted hERG1 blocking affinities [18–22]. Recently, cryo-EM
structure of hERG1 in open-state is also deposited to the pro-
tein data bank (PDB) by Rod MacKinnon group [23]. Therefore, we
also repeated hERG-related docking studies on the cryo-EM struc-
ture.
The aim of this study is to investigate the molecular mecha-
nisms of new generation potent AT1 antagonists designed in silico
and derive novel potent multi multi-functional anti-hypertensive
molecules having low side effects.
2. Methods
2.1. Protein preparation
Since the human AT1 crystal structure reported after the ini-
tiation of this project [24] all molecular modeling studies (i.e.,
protein preparation, molecular docking, and molecular dynamics
(MD) simulations) were also performed on homology model struc-
ture of AT1 together with the crystal structure. Homology model of
the AT1 [25] was constructed based on templates of the resolved
bovine rhodopsin structure [26]. AT1 homology model is refined via
short MD simulations and then used in the docking studies. Zhang
et al. published the first antagonist-bound AT1 X-ray structure in
April 2015 with 2.9 Å resolution, (PDB ID: 4YAY) [24]. In addition,
the same research group revealed AT1 structure binding with an
inverse agonist olmesartan at 2.8 Å resolution, (PDB ID: 4ZUD) [27].
This structure (4ZUD) does not involve 8-TM short helix, which is
essential in structural dynamics of GPCRs. Hence, in our simula-
tions we used full TM crystallized structure (4YAY) since it includes
8-TM short helix as well as Kellici et al. claimed that bioactive con-
formation of olmesartan in the two co-crystallized structures did
not show different results [28]. The X-ray crystal structure of the
AT1 receptor (PDB ID, 4YAY) and hERG1 open state cryo-EM (PDB
ID, 5VA1) structures were retrieved from RCSB Protein Databank.
Downloaded crystal structure and 3D homology model structure
were prepared and minimized for the further steps using Protein
Preparation module of Maestro molecular modeling suite [29]. The
stabilizer residues from D1002 to L1106 in AT1 crystal structure
were deleted before optimization and minimization stages. Bond
orders were assigned, hydrogen atoms were added, disulfide bonds
were created, missing side chains were added, loops were fixed

Fig. 1. Superimposition of the top Glide/HTVS (yellow) and Glide/SP (purple) docking poses of ZD7 ligand conformations in binding pocket. (left: Yellow colored docking
pose indicates Glide/HTVS top-docking result, cyan-colored conformation indicates conformer of ZD7 at the crystal structure of AT1; right: purple-colored conformation of
ZD7 indicates the conformation of ligand from Glide/SP docking pose and cyan-colored conformation indicates conformer of ZD7 at the crystal structure of AT1).
 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117 105

Table 1
Comparison of top-docking poses of the ligand ZD7 (co-crystallized ligand) with its conformation at the crystal structure of AT1. Table shows RMSD values using heavy atoms
of ligand.

Molecular docking software

Glide Gold

Glide/HTVS (High Glide/SP Glide/XP (Extra IFD (Induced Fit GoldScore Fitness ChemScore DG
Throughput Virtual (Standard Precision) Docking)
Screening) Precision)

RMSD (Å) 2.88 0.70 0.68 0.70 0.66 0.83

using Prime module of Maestro as preprocess of refinement. pKa Protonation states of ligands were assigned by Epik and pH was set
predictions and optimizations were implemented at the physio- to neutral pH 7.0 [35,36].
logical pH 7.4 using PROPKA [30,31]. Energy minimization was
performed by OPLS2005 force field [32,33].
2.3. Molecular docking

2.2. Ligand preparation
18 million “drug-like” small molecules were retrieved from ZINC
database [34]. Commercially available sartan molecules (losar-
tan, valsartan, candesartan, eprosartan, olmesartan, irbesartan,
telmisartan, olmesartan-medoxomil, azilsartan-medoxomil, and
candesartan-cilexetil) were downloaded from ChEMBL database.
Oxazolone and imidazolone molecules were initially sketched in 2D
Sketcher in Maestro molecular modeling package. De novo studies
of telmisartan, azilsartan medoxomil, imidazolone, and oxazolone
molecules were produced based on previous ligand interaction
maps with 3464 small molecule fragments which were designed
and prepared by our group. Structural optimizations of the ligands
were carried out using Ligprep with OPLS2005 [32,33] force field.
Molecular docking studies were performed on Glide [37–39]
and GOLD [40] docking programs. Grid maps were produced in
known active sites of AT1 homology model [25,26,41] and pre-
pared and refined AT1-X-ray crystal structure [24]. Ligands and
rotatable amino acid bonds of the active site of the target proteins
were treated as flexible in docking protocols. Ligands screening
procedure was based on hierarchical screening approach; high
throughput virtual screening (HTVS), standard precision (SP), extra
precision (XP), induced fit docking (IFD), and Quantum-Polarized
Ligand Docking (QPLD) protocols [37–39] of Maestro molecular
modeling package were applied to 18 million ZINC compounds and
selected sartan compounds, imidazolone, and oxazolone derived
molecules, respectively. Ligand molecules were filtered by dock-
ing algorithms with considering only molecules that have up to

Table 2
Docking simulation analysis with the selected 5 hit compounds from 18 million small molecules ZINC database in AT1 receptor crystal structure (4YAY, PDB ID).

Hit Molecules from 18 Million Small Molecules Bank Glide

HTVS SP XP IFD QPLD

−8.02 −9.65 −11.42 −15.19 −11.48

−8.59 −9.23 −11.06 −14.57 −11.21

−8.57 −9.18 −11.33 −14.50 −11.32

−9.06 −9.75 −11.03 −14.18 −11.10

−8.02 −9.02 −11.93 −13.83 −12.49

106 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

Fig. 2. 3D and 2D ligand interaction diagrams of top IFD pose of oxazolone fragment329 molecule at the binding pocket of AT1 crystal structure.

Fig. 3. 3D and 2D ligand interactions of top scored ZINC database molecule (ZINC19834701) at the binding pocket of the AT1 crystal structure. 3D and 2D interactions of the
top scored ZINC molecule in AT1 homology model structure is indicated in Supplementary Material Fig. S2.

2 kcal/mol lower docking scores than top-docking scored molecule
when its filtered from Glide/HTVS to Glide/SP, from Glide/SP to
Glide/XP, and from Glide/XP to IFD. Glide/QPLD and GOLD dock-
ing was performed for selected filtered ligands having high docking
scores as a result of the filtering process. Ligand binding energies
were predicted by GoldScore Fitness and ChemScore DG scoring
functions in GOLD docking software. Eight amino acid residues in
active sites of the target structures (Tyr35, Phe77, Trp84, Ser109,
Leu112, Arg167, Lys199, His256) were treated as flexible by uti-
lizing the rotamer library implemented in the GOLD program.
Default settings were used for population (population size: 100)
and genetic operations (number of operations: 100000) steps.
Search efficiency was set to its maximum value (200%) and 20
docking poses per ligand were generated in order to increase the
reliability of the derived binding modes throughout the docking
simulations. Hierarchical screening steps are given in Scheme 1.
2.4. Molecular dynamics (MD) simulations
MD simulations are essential to examine the structural and
dynamical profile of ligand/protein interactions qualitatively and
quantitatively in atomistic detail. Simulations were performed for
apo (ligand-free) and holo (with ligand) forms of the target struc-
tures. The initial protein-ligand complex structures used in MD
simulations were selected according to their ligand interaction
profile with high docking scores obtained from the IFD protocol.
Each system box generated in rectangular form and includes 256
DPPC (dipalmitoylphosphatidylcholine) molecules (128 lipids for
both upper and lower leaflets) surrounding the protein-ligand com-
plex. Systems were solvated with TIP3P water molecules of 20 Å
thickness. System’s neutralization was done by Monte-Carlo ion
placing method with addition of 0.15 M NaCl. CHARMM-GUI mem-
brane builder interface was used to generate system inputs [42–44].
Receptor orientation in lipid bilayer was specified according to the
output of orientations of proteins in membranes (OPM) database
[45,46]. MD simulations were performed with Gromacs 4.6.5 soft-
ware package [47–49]. A prior energy minimization was applied to
the full system using the steepest-descent (SD) integrator for 5000
steps with the initial step size of 0.01 Å (the energy minimization
tolerance was set to 1000 kJ/mol). The systems were then equili-
brated for 5 ns MD simulations in a total of 6 different steps using
Berendsen barostat and thermostat [50] algorithms. Bond lengths
were constrained using LINCS algorithm and Particle Mesh Ewald
(PME) method [51] was used to calculate long-range electrostatic
interactions. Cut-off distances for the calculation of Coulomb and
van der Waals interactions were both 12 Å. 100 ns and 0.5 ␮s pro-
duction runs were performed for different systems with a time
step of 2.0 fs using Nose-Hoover thermostat [52] and Parrinello-
 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117
Fig. 4. Top-docking poses of telmisartan at the binding pocket of hERG1 pore domain. (top) hERG1 3D model; (bottom) hERG1 cryo-EM structure.
Rahman barostat. Simulations were run in the NPT ensemble at
310 K and 1 bar with periodic boundary conditions (PBC). Visual-
ization of the dynamics trajectories was performed with the VMD
software package.
2.5. Molecular mechanics generalized born solvation (MM-GBSA)
calculations
MM/GBSA method were used to calculate the ligand bind-
ing affinities by combining molecular mechanics and continuum
solvent models [53]. Prime − MM-GBSA module is used for protein-
ligand free energy analysis. OPLS2005 force field and VSGB 2.0
solvation model is used for calculations.
2.6. AT1–AT1 homodimerization
Growing evidences have suggested that GPCRs may exist in
oligomer forms [54,55]. A recent study has suggested that AT1
receptor interacts with 4th, 5th, 6th, and 7th-TM regions to form
homodimer structure [56]. There is no available crystal structure
for AT1 dimer protein. Thus, available GPCR dimer structure as
topological template which is constructed previously by our group
that has experimental validations was used in this study [57]. Con-
structed homodimer model structures were simulated in apo form
and with a standard AT1 inhibitor- telmisartan.
107
3. Results and discussion
The co-crystallized ligand (ZD7) was re-docked to the AT1 crys-
tal structure for predicting success of molecular docking protocols
to find correct binding mode of ligands at active site. Predicted
docking poses by different docking algorithms were compared with
co-crystallized complex structure (Fig. 1) and root mean square
deviation (RMSD) values were compared (Table 1). RMSD values
below 2 Å may indicate that the protocol is able to predict the cor-
rect binding mode. Predicted docking conformations were below
1.0 Å except Glide/HTVS. Although Glide/HTVS has slightly larger
RMSD value (2.88 Å), it predicts roughly correct placement of lig-
and’s fragments at the binding pocket of the target. Thus, the
docking programs are capable of predicting the bioactive confor-
mation of the ligands in the AT1 receptor binding pocket.
Sartan molecules are known commercially available antagonists
of AT1 receptor. In the current study, available sartan molecules
(Table S1, Supplementary Materials) were considered as positive
controls for molecular docking studies. Glide docking protocols
were implemented for sartans both in AT1 crystal and homology
model structures and results were listed in Tables S1 and S2 at the
Supplementary Materials.
Imidazolone and oxazolone molecules were designed as AT1
antagonists. Relying on the preliminary molecular screening and
de novo design studies, imidazolone and oxazolone molecules
were derivatized from different enumeration sites using a small-
fragment library consisting of 3464 fragments that were designed
and prepared by our group. (see Supplementary Materials, Fig. S1)
More than 50000 oxazolone and imidazolone based compounds
108 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

Fig. 5. (top) RMSD-time graph of for the C␣ atoms of the AT1 receptor (PDB ID: 4YAY) complexed with molecules represented as different colors. Average RMSD val-
ues throughout the MD simulation: apo = 2.35, ZD7 = 2.65, Telmisartan = 2.51, Azilsartan Medoxomil = 3.14, Oxazolone-fragment329 = 3.32, Telmisartan-fragment310 = 2.62,
Azilsartan Medoxomil-fragment131 = 2.62. (bottom) Corresponding plots of telmisartan and oxazolone-fragment329 compounds were compared separately for clarity.
Oxazolone-fragment329 molecule is clearly indicated in Fig. S3 of the Supplementary Materials.

were derived and screened at the binding pockets of both AT1 crys-
tal and 3D model structures using Glide modules (HTVS, SP, XP,
IFD, and QPLD) (see Supplementary Materials Tables S3 and S4).
Top-scored (Glide/IFD) molecules among derived imidazolone and
oxazolone compounds were also docked by GOLD molecular dock-
ing software [40] (Supplementary Materials, Table S3). In addition,
the same de novo derivatization and molecular docking protocols
were also implemented for selected sartan molecules (Supplemen-
tary Materials, Table S5).
Oxazolone-fragment329 were the one of the highest scoring
molecules in both GOLD and Maestro/Glide softwares. The 3D and
2D ligand interaction diagrams from top-scored IFD docking pose
of compound oxazolone-fragment329 are represented in Fig. 2. The
oxazolone-fragment329 molecule interacts with crucial amino acid
residues at the binding cavity of AT1 such as Phe77, Trp84, His256,
Arg167, Ile288 and Tyr292. While compound constructed ␲-␲
stacking interactions with Phe77, Trp84 and His256; hydrogen-
bonding interactions occur from Arg167, Ile288 (backbone), and
Tyr292. Molecule also stays in the vicinity of other hydrophobic
(i.e., Val108, Met284, Leu81) and polar (i.e., Gln267, Thr260, Ser109)
residues. The results were also compared with key amino acids as
at the binding cavity as well as docking scores of known AT1 sartans
such as telmisartan and azilsartan.
Large ligand databases consisting of small molecules were
screened with Glide docking software on AT1 crystal structure [24]
and also AT1 3D homology model [25,26]. For this purpose, approx-
imately 18 million “drug-like” small molecules were downloaded
from ZINC database [34] and prepared with Ligprep module in
physiological pH (pH 7.4) for molecular docking studies [35,36].
Obtained molecules were then subjected to Glide/HTVS, Glide/SP,
Glide/XP, and Glide/IFD procedures [37–39]. Quantum polarized
ligand docking-QPLD was also performed for the selected top-5
molecules with high docking scores at the AT1 binding pocket from
virtual screening and these molecules were listed in Table 2 and
 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117 109

Fig. 6. RMSD-time graph for the ligands which are used in AT1 crystal structure receptor/ligands MD simulations (all-atoms of ligands were considered).

Supplementary Material Table S6. The 3D and 2D ligand interac- Table 3

Docking scores of selected molecules in hERG1 potassium ion channel’s central
tion diagrams of the top-docking scored molecules selected from
cavity.
18 million ZINC database compounds screened in crystal and AT1
homology model structures were shown in Fig. 3 and Supplemen- Compound hERG1 (5VA1) hERG1 3D
Glide/SP Homology Model
tary Materials, Figure S2. The crucial amino acids considered to
Glide/SP
be important for the hit molecule (ZINC19834701) at the bind-
ing pocket of AT1 crystal structure were following: H-bonding, Telmisartan −9.43 −9.25
Azilsartan Medoxomil-fragment131 −8.87 −9.61
Lys199 and Tyr35; ␲-␲ stacking, Trp84; polar, Thr88; charged,
Losartan −7.99 −7.01
Arg167; hydrophobic, Phe77, Leu81, Ile288, and Ala163. The cor- Olmesartan Medoxomil −7.48 −7.17
responding residues for the selected top-scored hit molecule Imidazolone-fragment415 −7.28 −10.41
(ZINC39383781) at the binding cavity of AT1 homology model Irbesartan −6.72 −6.56
Candesartan Cilexetil −6.52 −8.86
were following: H-bonding, Ser105, Cys180 (backbone); ␲-cation,
Azilsartan Medoxomil −6.45 −9.50
Trp253; hydrophobic, Ala181, Val179, Phe171, Ile288, Ala291, Eprosartan −6.25 −6.62
Val108, Leu81, Phe204, Val116, Phe208, Leu112 and Phe182. Oxazolone-fragment329 −6.23 −6.98
Promising molecules were then docked to open state hERG1 ZINC93960476 −6.16 −6.31
cryo-EM structure and hERG1 3D homology models. According ZINC19834701 −5.59 −5.59
ZINC02483401 −5.99 −5.62
to these docking scores, the hERG blockage activity of suggested
ZINC04324564 −5.98 −5.34
molecule was found to have lower or similar to the commercially Telmisartan-fragment310 −5.69 −9.74
available antihypertensive drugs (Fig. 4 and Table 3). While Ser624, Olmesartan −5.38 −5.38
Tyr652, and Gln664 were found as critical amino acids for telmis- ZINC08877909 −5.26 −7.68
Candesartan −5.20 −6.02
artan using cryo-EM structure, corresponding amino acids were
Valsartan −4.19 −5.19
Ser624, Tyr652, and Phe656 using hERG1 3D homology model
structure.

were also applied to AT1 3D homology model structure. RMSD-time

3.1. Analysis of the activation mechanism of AT1 receptor via
molecular dynamics (MD) simulations, ligand binding studies and
atomistic models
Oxazolone and imidazolone derivatives, known sartan
molecules, and generated sartan derivatives have been selected
based on the lowest binding energy scores (i.e., top-docking poses)
and submitted to the MD simulations using crystal structure [24]
(both monomer and dimer forms) and AT1 receptor homology
model [25]. Known sartan molecules were chosen in order to make
comparison between the positive controls and computer-based
designed oxazolone and imidazolone derivatives.
RMSD calculations for the C␣ atoms of the AT1 receptor dur-
ing MD simulations were performed for studied systems. (Fig. 5)
It is observed that all systems have RMSD values smaller than
4 Å. RMSD-time graph for oxazolone-fragment329 is clearly indi-
cated at Fig. S3 of the Supplementary Materials. Same MD protocols
graph for the molecules that have the high docking scores on AT1
model is shown in Fig. S4 of the Supplementary Materials. It can
be clearly seen that RMSD values of simulations with AT1 homol-
ogy model are determined to be higher (i.e., apo-form and system
complexed with imidazolone-fragment335 have higher RMSD val-
ues (>5 Å)) than the trajectories obtained from simulations of AT1
crystal structure. For example, while ZD7-complexed target has
the RMSD values around 3 Å at the crystal structure, corresponding
value was around 5 Å for homology model of AT1.
RMSD-time graphs for ligands (all atoms) are shown in Fig. 6
(for trajectories obtained from simulations of AT1 crystal struc-
ture). Corresponding figure for MD simulations with AT1 homology
model has been shown at Fig. S5 of the Supplementary Materi-
als. Results show that simulations that are performed using crystal
structures have smaller RMSD than results obtained from simula-
tions with AT1 homology model. Moreover, its shown that ligands
are stable at the binding pocket throughout the MD simulations.
110 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117
Fig. 7. (top) RMSF graph of the AT1 crystal structure with apo form and complexed with different ligands in 100 ns MD simulations. (bottom) Corresponding plots of
telmisartan and oxazolone-fragment329 compounds were compared separately for clarity.
The most fluctuated sites of the protein during the simulations
are indicated on root mean square fluctuation (RMSF) graphs
(Fig. 7). According to the RMSF results, TM domains (residue num-
bers, 27–55 (TM1); 62–89 (TM2); 99–130 (TM3); 138–164 (TM4);
190–224 (TM5); 236–268 (TM6); 274–304 (TM7)) are observed as
rigid and have lower RMSF values or in other words they have less
fluctuation compared to the loop and linker regions, as expected.
TM5 region showed higher fluctuation compared to other TM
domains. Especially 225–235 TM5-TM6 long loop region showed
higher fluctuations in all studied systems as expected. Interestingly,
8th-short helix at the AT1 showed very mobile behavior throughout
the MD simulations. Fig. 7 (bottom) suggests that telmisartan and
oxazolone-fragment329 bound AT1 systems have similar RMSD
plots at TM-domains, however telmisartan-bound target shows
higher fluctuations especially at TM5-TM6 loop regions. It can also
be observed that the homology model of AT1 receptor was observed
to be more mobile compared to the crystal structure (Fig. S6, Sup-
plementary Materials).
Conformational changes of ligand and protein during simula-
tions were also visualized. The changes occur within the systems
of AT1 crystal structure complexed with oxazolone-fragment329,
ZD7, and telmisartan molecules were given in Fig. 8, and Figs.
S7 and S8 at the Supplementary Materials, respectively. While
Arg167 constructs an H-bond from carbonyl oxygen at the oxa-
zolone ring (top-docking pose), throughout the simulations this
bond disappear and a new H-bond occur between Arg167 and
tetrazole ring fragment (Fig. 8). By this way, tetrazole ring also con-
structs another H-bonding interaction with another crucial amino
acid Lys199. With this bioactive conformation of ligand, oxazolone-
frgament329 is able to form a ␲-cation interaction with the central
aromatic ring of the compound. (Fig. 8) When the trajectory frames
of telmisartan investigated, it can be seen that while Arg167 keeps
its interaction pairs, Lys199 also forms another hydrogen bonding
interactions from carboxyl group of the ligand throughout simula-
tions. (Supplementary Materials, Fig. S8)
In order to see the effect of simulation time restrictions
on the studied systems, telmisartan and oxazolone-fragment329
molecules complexed with the AT1 systems, the simulations were
extended up to 0.5 ␮s. (Fig. 9 and Fig. S9 at the Supplementary
Materials). As its clear from corresponding figures, extension of
simulation time did not lead different results in terms of compar-
ison of RMSD and RMSF graphs of C␣ atoms of the AT1 receptor.
 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117
Fig. 8. Conformational changes of oxazolone frag329 molecule in AT1 crystal receptor structure during simulation time (t = 0, red; t = 100 ns, blue). Ligand-protein interaction
diagrams on the initial and at the final trajectory frames of MD simulation are indicated with arrows in corresponding colors. (For interpretation of the references to colour
in this figure legend, the reader is referred to the web version of this article.)
Table 4
Molecular Mechanics Generalized Born Surface Area (MM/GBSA) method is used for
free energy (G) analyses in systems.
Selected Hit Molecules MM/GBSA (kcal/mol)
ZD7 −105.59
Telmisartan −115.70
Azilsartan Medoxomil −111.28
Telmisartan-fragment310 −132.64
Azilsartan Medoxomil-fragment131 −130.78
Oxazolone-fragment329 −91.42
However, the ligand RMSD of oxazolone-fragment329 showed an
interesting behavior, RMSD value of ligand increased from 0.5-
1.0 Å to around 3.0 Å between 250 and 375 ns time-scale, but it
decreased again to near 1.0-1.5 Å at the rest of the simulation. In
contrast, ligand RMSD of telmisartan was very stable (around 0.5-
1.0 Å) throughout simulations. Thus, it can be interpreted that while
docking pose of telmisartan looks quite stable (in terms of struc-
tural fluctuations) throughout simulations, initial conformation of
oxazolone-fragment329 is re-optimized with additional interac-
tions at the binding cavity of AT1 receptor.
Molecular Mechanics Generalized Born Surface Area
(MM/GBSA) calculations were implemented for top-docking
scored sartan and oxazolone derivative molecules (Table 4).
Results show that especially telmisartan fragment310 and azil-
sartanmedoxomil fragment131 molecules have better interaction
energy profiles compared to corresponding wild types.
As its indicated in Methods chapter, it has been suggested that
GPCRs can exist in oligomer forms. Thus, homodimer structures
of AT1 have been designed. Constructed homodimer structure can
be found in Fig. S10 at the Supplementary Materials. Constructed
dimer AT1 protein was used as target in docking simulations and
apo form and top-docking scored poses have been used in MD sim-
ulations. Figs. S11 and S12 at the Supplementary Materials show
RMSD and RMSF graphs. Corresponding figures for holo forms have
been represented in Figs. S13 and S14, respectively. When it is com-
111
pared to monomer, telmisartan-bound system and apo-form of C␣
atoms of the AT1-dimer receptor showed similar behavior. How-
ever, when the RMSDs of ligand were compared, telmisartan at the
dimer structure represented slightly smaller RMSD values com-
pared to the corresponding monomer plot. Comparison of RMSF
plots show that especially loop regions become less mobile at the
dimer structure (i.e., while TM5-TM6 loop region has more than 6 Å
RMSF values at the monomer, corresponding value is less than 3 Å
at the dimer structure).
3.2. MetaCore/MetaDrug applications
Moreover, the selected hit molecules as well as one of the pos-
itive control molecules (telmisartan) were put through further
investigation regarding their therapeutic activity as well phar-
macokinetic properties using MetaCore/MetaDrug comprehensive
systems biology analysis suite of Clarivate Analytics with the help
of available ADME, disease and toxicity QSAR models. MetaCore is
an integrated software suite for functional analysis of Next Gen-
eration Sequencing, gene expression, CNV, metabolic, proteomics,
microRNA, and screening data. MetaCore is based on a high-quality,
manually-curated database of molecular interactions, molecu-
lar pathways, gene-disease associations, chemical metabolism
and toxicity information. MetaCore incorporates extensive man-
ually curated information on biological effects of small molecule
compounds. Table 5 shows results of ADME QSAR models, pre-
diction of therapeutic activities, and prediction of toxic effects.
Antihypertensive activity prediction of tested molecules with
the help of used QSAR models at the MetaCore/MetaDrug suite
(Training set consists of approved drugs; training set N = 554,
test set N = 111, sensitivity = 0.89, specificity = 0.81, accuracy = 0.85,
MCC = 0.70), shows that proposed compounds Telmisartan-
fragment310, Azilsartan medoxomil-fragment131, Imidazolone-
fragment415, ZINC04324564, and ZINC19834701 may have
antihypertensive effects since they have same or higher values than
112 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

Fig. 9. RMSD and RMSF-time graphs for the C␣ atoms of the AT1 crystal structure complexed with telmisartan and oxazolone frag329 molecules.

cutoff value of 0.5. These compounds were also checked with 26-
different toxicity criteria using toxicity QSAR models and almost
no-toxicity were found for these compounds (please see Table 5).
For example, while telmisartan shows predicted slight inducing
cardiotoxicity (0.53; cutoff is 0.5. Values higher than 0.5 indicate
potentially risky compounds), ZINC04324564 and imidazolone-
fragment415 shows no any predicted toxicity against these 26
QSAR toxicity models. Cardiotoxicity risks of ZINC04324564 and
imidazolone-fragment415 molecules were predicted as 0.21 and
0.33, respectively. Indeed, these data fits well with the docking
scores of selected molecules in hERG1 channel. (Table 3) Dock-
ing scores of telmisartan at the pore domains of both hERG1
cryo-EM and model structures are found high. Tu et al. [58]
measured the IC50 value of telmisartan as 24 ␮M at the hERG
channel and they found that blockage is voltage- and concentra-
tion dependent. Corresponding values for hit molecules such as
ZINC04324564 and ZINC19834701 have smaller docking score val-
ues at the pore domains of the hERG1 channel. ADME QSAR models
of hit molecules (Table 5) also give interesting results. While human
serum protein binding percentage and affinity to human serum
albumin (log value of the retention time) of telmisartan show that it
has high affinity to serum albumin (86.49% (cutoff is 50%) and 0.28
(cutoff is 0), respectively); one of the hit molecules, ZINC19834701
has very low corresponding values (48.17%, and −0.40, respec-
tively).
Table 5
MetaCore/MetaDrug Applications. (top) ADME QSAR models; (mid) Prediction of therapeutic activity; (bottom) Prediction of toxic effects. (Values within the paranthesis show Tanimato Prioritization (TP) values which has values
between 0 and 1 and it is the maximal Tanimato coefficient calculated for all of the molecules that are included in the training set. TP is similar to the most similar hit of the QSAR training set.).

ADME QSAR Models

Telmisartan Telmisartan frag310 Azilsartan Medoxomil frag131 Oxazolone-frag329 Imidazolone frag415 ZINC93960476 ZINC02483401 ZINC04324564 ZINC08877909 ZINC19834701

BBB, log ratio (1) −0.97 −0.75 −0.84 (41.42) −0.38 (37.92) −0.51 (33.17) −0.98 (32.92) −0.70 (31.58) −0.40 (39.29) −0.94 (54.05) −0.63 (40.00)
G-LogP (2) 4.15 4.20 3.05 2.88 2.74 3.21 2.02 3.07 3.14 1.81
Prot-bind, % (3) 86.49 (42.16) 83.11 (45.48) 86.63 (33.64) 92.63 (34.99) 90.36 (34.66) 85.54 (40.26) 71.69 (31.58) 62.39 (47.46) 81.57 (54.05) 48.17 (43.55)
Prot-bind, Log t (4) 0.28 (41.47) 0.27 (42.70) 0.22 (41.42) 0.08 (37.92) 0.07 (33.17) −0.39 (32.92) −0.41 (31.58) −0.04 (39.29) 0.08 (54.05) −0.40 (40.00)
WSol, log mg/L (5) 1.64 1.81 1.63 1.46 1.87 2.16 0.60 1.87 1.16 2.57

Prediction of Therapeutic Activity

Telmisartan Telmisartan frag310 Azilsartan Medoxomil frag131 Oxazolone-frag329 Imidazolone frag415 ZINC93960476 ZINC02483401 ZINC04324564 ZINC08877909 ZINC19834701

B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

Allergy (6) 0.76 (53.54) 0.20 (46.49) 0.45 (36.30) 0.21 (37.12) 0.13 (34.22) 0.68 (46.41) 0.50 (32.83) 0.32 (37.61) 0.48 (47.57) 0.31 (34.30)
Alzheimer (7) 0.69 (38.40) 0.55 (50.47) 0.67 (39.92) 0.62 (32.55) 0.47 (37.53) 0.30 (38.46) 0.48 (33.97) 0.67 (45.97) 0.33 (46.86) 0.63 (37.31)
Angina (8) 0.25 (55.56) 83.11 (45.48) 0.25 (54.35) 0.21 (37.12) 0.21 (35.04) 0.04 (46.11) 0.06 (32.15) 0.17 (43.22) 0.18 (45.26) 0.17 (42.65)
Arthiritis (9) 0.73 (49.62) 0.96 (46.67) 0.82 (35.25) 0.84 (32.16) 0.92 (35.04) 0.59 (37.50) 0.70 (40.06) 0.61 (48.28) 0.52 (46.86) 0.55 (37.31)
Asthma (10) 0.74 (41.22) 0.43 (54.03) 0.45 (36.65) 0.57 (36.73) 0.67 (37.53) 0.51 (45.14) 0.72 (40.06) 0.68 (48.28) 0.45 (46.86) 0.56 (36.92)
Bacterial (11) 0.31 (43.56) 0.35 (51.88) 0.60 (42.16) 0.62 (38.88) 0.32 (37.56) 0.23 (55.78) 0.41 (34.06) 0.08 (54.04) 0.24 (49.64) 0.25 (49.41)
Cancer (12) 0.69 (43.02) 0.67 (46.62) 0.60 (42.30) 0.63 (39.28) 0.66 (38.33) 0.41 (48.57) 0.57 (31.72) 0.28 (54.04) 0.27 (51.79) 0.31 (49.41)
Depression (13) 0.36 (41.98) 0.13 (49.29) 0.07 (35.28) 0.36 (33.89) 0.31 (34.84) 0.28 (55.78) 0.14 (30.41) 0.48 (40.72) 0.20 (38.98) 0.47 (39.17)
Diabetes (14) 0.34 (34.22) 0.26 (46.49) 0.09 (36.15) 0.28 (31.90) 0.34 (35.32) 0.59 (45.99) 0.90 (34.30) 0.41 (41.04) 0.20 (38.78) 0.52 (38.86)
HIV(15) 0.54 (41.22) 0.42 (54.03) 0.38 (37.40) 0.48 (36.73) 0.76 (37.53) 0.54 (45.14) 0.74 (40.06) 0.69 (48.28) 0.36 (46.86) 0.56 (38.24)
Heart Failure (16) 0.61 (55.56) 0.30 (52.26) 0.59 (54.35) 0.77 (37.12) 0.85 (32.19) 0.19 (44.00) 0.88 (32.53) 0.79 (41.53) 0.69 (37.07) 0.69 (34.96)
Hyperlipidemia (17) 0.25 (40.81) 0.37 (40.29) 0.34 (36.69) 0.66 (35.03) 0.15 (36.25) 0.58 (42.95) 0.63 (38.70) 0.69 (41.79) 0.86 (44.27) 0.64 (41.39)
Hypertension (18) 0.93 (100.00) 0.50 (74.90) 0.65 (56.11) 0.47 (38.76) 0.61 (45.43) 0.27 (46.11) 0.47 (33.61) 0.77 (46.24) 0.48 (39.14) 0.74 (44.69)
Inflammation (19) 0.22 (42.59) 0.13 (39.90) 0.23 (41.86) 0.58 (36.21) 0.28 (36.25) 0.76 (48.57) 0.32 (32.83) 0.27 (44.78) 0.39 (54.47) 0.24 (44.87)
Migraine (20) 0.42 (41.22) 0.29 (54.12) 0.42 (36.65) 0.56 (36.73) 0.70 (37.53) 0.69 (45.14) 0.56 (34.30) 0.70 (48.28) 0.30 (46.86) 0.55 (38.86)
Mycosis (21) 0.41 (52.91) 0.61 (45.74) 0.57 (34.64) 0.52 (33.65) 0.35 (36.25) 0.14 (46.11) 0.40 (29.13) 0.22 (41.04) 0.36 (37.50) 0.20 (33.33)
Obesity (22) 0.99 (41.22) 0.98 (54.12) 1.00 (35.31) 0.96 (36.73) 0.96 (36.09) 0.77 (42.51) 0.86 (40.06) 0.92 (48.28) 0.95 (46.86) 0.71 (38.24)
Osteoporosis (23) 0.33 (55.56) 0.70 (54.12) 0.21 (54.35) 0.71 (36.73) 0.60 (36.76) 0.49 (42.51) 0.21 (40.06) 0.42 (48.28) 0.52 (46.86) 0.17 (38.24)
Pain (24) 0.41 (49.60) 0.24 (47.00) 0.08 (41.46) 0.48 (36.21) 0.27 (37.20) 0.66 (47.71) 0.13 (31.04) 0.60 (51.30) 0.37 (39.11) 0.43 (53.25)
Parkinson (25) 0.36 (39.77) 0.41 (52.61) 0.52 (39.92) 0.35 (34.20) 0.30 (34.84) 0.16 (48.57) 0.30 (27.46) 0.31 (43.87) 0.19 (38.94) 0.51 (39.81)
Psoriasis (26) 0.32 (41.34) 0.31 (52.26) 0.25 (39.92) 0.27 (33.23) 0.38 (34.66) 0.30 (44.12) 0.28 (29.76) 0.46 (39.71) 0.26 (43.13) 0.73 (33.67)
Schizophrenia (27) 0.36 (53.54) 0.28 (48.80) 0.24 (41.65) 0.30 (37.39) 0.51 (36.16) 0.17 (42.46) 0.15 (35.28) 0.14 (42.63) 0.49 (51.53) 0.12 (41.63)
Skin Diseases (28) 0.13 (41.47) 0.17 (51.88) 0.24 (41.42) 0.44 (37.97) 0.09 (36.39) 0.70 (45.99) 0.54 (30.13) 0.10 (36.94) 0.42 (49.50) 0.10 (34.89)
Trombosis (29) 0.30 (49.52) 0.30 (47.00) 0.16 (41.65) 0.21 (37.39) 0.40 (37.56) 0.14 (38.50) 0.28 (33.61) 0.30 (44.78) 0.28 (53.05) 0.25 (44.87)
Viral (30) 0.48 (37.07) 0.37 (52.26) 0.48 (32.68) 0.31 (32.42) 0.43 (34.35) 0.35 (44.00) 0.49 (33.24) 0.32 (45.97) 0.26 (37.03) 0.39 (41.63)

Prediction of Toxic Effects

Telmisartan Telmisartan frag310 Azilsartan Medoxomil frag131 Oxazolone-frag329 Imidazolone frag415 ZINC93960476 ZINC02483401 ZINC04324564 ZINC08877909 ZINC19834701

AMES (31) 0.44 (31.15) 0.35 (31.76) 0.44 (31.15) 0.23 (33.97) 0.37 (31.42) 0.47 (44.03) 0.16 (32.68) 0.20 (39.32) 0.23 (46.52) 0.32 (36.92)
Anemia (32) 0.06 (39.16) 0.51 (46.59) 0.06 (39.16) 0.16 (34.87) 0.35 (35.41) 0.33 (42.46) 0.05 (33.52) 0.25 (47.46) 0.16 (41.42) 0.33 (43.55)
Carcinogenicity (33) 0.03 (42.16) 0.04 (51.70) 0.03 (42.16) 0.20 (38.88) 0.08 (36.84) 0.34 (50.81) 0.25 (33.80) 0.13 (45.97) 0.09 (55.00) 0.18 (41.80)
Carci. Mouse (F)(34) 0.03 (42.16) 0.07 (51.70) 0.03 (42.16) 0.43 (38.88) 0.33 (35.18) 0.36 (50.81) 0.52 (38.37) 0.05 (43.46) 0.15 (45.26) 0.13 (41.80)
Carc. Mouse(M)(35) 0.02 (42.01) 0.10 (51.70) 0.02 (42.01) 0.35 (37.97) 0.22 (35.87) 0.49 (50.81) 0.42 (38.37) 0.05 (43.46) 0.35 (55.00) 0.15 (41.80)
Carcin. Rat (F) (36) 0.04 (42.16) 0.04 (51.70) 0.04 (42.16) 0.22 (38.88) 0.17 (37.45) 0.20 (44.03) 0.38 (38.37) 0.07 (45.97) 0.14 (46.39) 0.26 (41.80)
Carcin. Rat (M) (37) 0.01 (42.16) 0.02 (51.70) 0.01 (42.16) 0.23 (38.88) 0.12 (37.45) 0.37 (50.81) 0.39 (38.37) 0.03 (45.97) 0.08 (46.52) 0.13 (41.80)
Cardiotoxicity (38) 0.53 (44.79) 0.66 (74.90) 0.53 (44.79) 0.27 (32.14) 0.33 (33.09) 0.13 (44.03) 0.12 (33.61) 0.21 (51.30) 0.16 (45.26) 0.17 (53.25)
Cytox −log GI50 (M) (39) 4.40 (42.42) 4.89 (51.93) 4.40 (42.42) 3.52 (65.33) 4.22 (37.75) 5.00 (42.24) 4.75 (31.68) 4.71 (48.37) 4.98 (54.92) 4.55 (47.76)
Epididymis Tox. (40) 0.19 (40.95) 0.38 (42.70) 0.19 (40.95) 0.45 (37.12) 0.13 (35.87) 0.64 (42.86) 0.34 (30.21) 0.02 (40.83) 0.08 (46.86) 0.02 (37.77)
Genotoxicity (41) 0.05 (42.16) 0.21 (51.70) 0.05 (42.16) 0.22 (38.88) 0.23 (35.87) 0.65 (50.81) 0.28 (33.64) 0.29 (47.46) 0.23 (45.26) 0.51 (43.55)
Hepatotoxicity (42) 0.11 (50.80) 0.13 (52.23) 0.11 (50.80) 0.24 (37.12) 0.20 (45.43) 0.41 (46.11) 0.25 (32.04) 0.22 (47.46) 0.23 (53.26) 0.26 (43.55)
Kidney Necrosis (43) 0.11 (34.12) 0.39 (38.29) 0.11 (34.12) 0.27 (31.17) 0.19 (32.77) 0.44 (40.22) 0.32 (33.52) 0.14 (45.73) 0.26 (42.74) 0.30 (41.06)
Kidney W.Gain (44) 0.04 (34.93) 0.05 (39.30) 0.04 (34.93) 0.09 (37.12) 0.04 (36.73) 0.69 (34.00) 0.19 (31.28) 0.04 (39.83) 0.18 (38.92) 0.06 (34.06)

113
 114
Liver Choles. (45) 0.04 (42.01) 0.09 (40.91) 0.04 (42.01) 0.35 (37.12) 0.09 (45.43) 0.68 (42.95) 0.26 (32.04) 0.39 (45.73) 0.56 (46.39) 0.37 (41.06)
Liver Lip. Accm (46) 0.47 (35.84) 0.37 (45.65) 0.47 (35.84) 0.28 (32.17) 0.25 (33.83) 0.59 (45.14) 0.33 (29.91) 0.33 (47.46) 0.23 (46.39) 0.43 (43.55)
Liver Necrosis (47) 0.38 (44.79) 0.27 (74.90) 0.38 (44.79) 0.73 (35.36) 0.27 (45.43) 0.97 (44.03) 0.73 (33.33) 0.11 (47.46) 0.27 (42.74) 0.61 (43.55)
Liver W. Gain (48) 0.14 (50.80) 0.19 (42.70) 0.14 (50.80) 0.73 (37.92) 0.12 (37.53) 0.97 (46.11) 0.74 (32.04) 0.06 (37.50) 0.07 (37.03) 0.13 (35.22)
MRTD (49) 0.18 (56.11) 0.08 (57.70) 0.18 (56.11) 0.76 (36.39) 0.34 (45.43) 0.65 (50.00) 0.34 (32.67) −0.00 (47.51) 0.17 (52.55) 0.31 (43.68)
Nasal Pathol. (50) 0.11 (34.36) 0.15 (42.86) 0.11 (34.36) 0.31 (33.87) 0.15 (35.87) 0.15 (37.36) 0.14 (31.82) 0.22 (44.78) 0.11 (36.12) 0.13 (44.87)
Nephron Injury (51) 0.06 (50.80) 0.45 (44.57) 0.06 (50.80) 0.50 (36.53) 0.18 (45.43) 0.50 (40.80) 0.42 (32.15) 0.07 (51.30) 0.18 (42.74) 0.26 (53.25)
Nephrotoxicity (52) 0.21 (50.80) 0.41 (74.90) 0.21 (50.80) 0.27 (37.92) 0.16 (38.33) 0.50 (38.95) 0.32 (33.61) 0.14 (53.62) 0.19 (42.74) 0.15 (46.61)
Neurotoxicity (53) 0.09 (40.13) 0.01 (50.16) 0.09 (40.13) 0.14 (37.12) 0.01 (34.40) 0.50 (45.14) 0.19 (30.57) 0.17 (53.62) 0.40 (38.92) 0.20 (46.61)
Pulmonary Tox.(54) 0.04 (41.10) 0.09 (42.70) 0.04 (41.10) 0.08 (37.97) 0.04 (45.43) 0.46 (42.54) 0.08 (34.30) 0.07 (54.04) 0.15 (41.45) 0.12 (49.41)
SkinSens, EC3 (55) 56.98 (26.43) 52.43 (26.99) 56.98 (26.43) 18.66 (51.10) 53.91 (28.80) 12.67 (38.35) 30.96 (25.43) 29.54 (30.56) 27.45 (31.72) 13.43 (28.57)
Testicular Tox.(56) 0.10 (40.95) 0.35 (42.86) 0.10 (40.95) 0.45 (37.12) 0.17 (35.87) 0.40 (42.86) 0.19 (38.70) 0.12 (44.78) 0.17 (54.47) 0.05 (44.87)
(1)
Blood brain barrier penetration model. The data is expressed as log values of the ratio of the metabolite concentrations in brain and plasma. Cutoff is −0.3. Larger values indicate that the metabolite is more likely to enter
the brain. Model description: N = 107, R2 = 0.89, RMSE = 0.26. (2) Lipophilicity, log of compound octanol-water distribution. Cutoffs are −0.4 to 5.6. Values greater than 5.6 correspond to overly hydrophobic compounds. Model
description: N = 13474, R2 = 0.95, RMSE = 0.21. (3) Human serum protein binding, %. Cutoff is 50%. A value of more than 95% is highly bound, less than 50% is a low binding metabolite. Model description: N = 265, R2 = 0.909,

B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

RMSE = 10.11. (4) Affinity to human serum albumin, log value of the retention time. Cutoff is 0. Positive values correspond to higher protein binding, negative values to lower protein binding. An acceptable level of binding is project
dependent. The model is based on retention times of compounds assayed by HPLC using an immobilized HSA column. Values are expressed as log values of the retention time. Model description: N = 95, R2 = 0.904, RMSE = 0.2.
(5)
Water solubility at 25 ◦ C, log mg/L. Cutoffs are from 2 to 4. An acceptable level of solubility is project dependent. Model description: N = 2871, 2 = 0.91, RMSE = 0.54. (6) Potential antiallergic activity. Cutoff is 0.5. Values higher
than 0.5 indicate potentially active compounds. Training set consists of approved drugs. Model description: Training set N = 258, Test set N = 47, Sensitivity = 0.87, Specificity = 0.88, Accuracy = 0.87, MCC = 0.74. (7) Potential activity
against Alzheimer’s disease. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity.
Model description: Training set N = 261, Test set N = 44, Sensitivity = 0.91, Specificity = 0.82, Accuracy = 0.86, MCC = 0.73. (8) Potential antianginal activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds.
Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set N = 546, Test set N = 95, Sensitivity = 0.90, Specificity = 0.93, Accuracy = 0.92,
MCC = 0.83. (9) Potential activity against arthritis. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds
with in vivo activity. Model description: Training set N = 460, Test set N = 77, Sensitivity = 0.98, Specificity = 0.94, Accuracy = 0.96, MCC = 0.92. (10) Potential activity against asthma. Cutoff is 0.5. Values higher than 0.5 indicate
potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set N = 366, Test set N = 63, Sensitivity = 0.92,
Specificity = 0.86, Accuracy = 0.89, MCC = 0.78. (11) Potential antibacterial activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs. Model description: Training
set N = 530, Test set N = 97, Sensitivity = 0.87, Specificity = 0.90, Accuracy = 0.89, MCC = 0.77. (12) Potential activity against cancer. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists
of approved drugs. Model description: Training set N = 886, Test set N = 167, Sensitivity = 0.89, Specificity = 0.83, Accuracy = 0.86, MCC = 0.72. (13) Potential activity against depression. Cutoff is 0.5. Values higher than 0.5 indicate
potentially active compounds. Training set consists of approved drugs. Model description: Training set N = 335, Test set N = 62, Sensitivity = 0.93, Specificity = 0.82, Accuracy = 0.87, MCC = 0.75. (14) Potential antidiabetic activity.
Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set
N = 195, Test set N = 34, Sensitivity = 0.85, Specificity = 0.93, Accuracy = 0.88, MCC = 0.77. (15) Potential activity against HIV. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved
drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set N = 491, Test set N = 80, Sensitivity = 0.80, Specificity = 0.86, Accuracy = 0.84, MCC = 0.67. (16) Potential activity
against heart failure. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs. Model description: Training set N = 204, Test set N = 33, Sensitivity = 0.78, Specificity = 0.87,
Accuracy = 0.82, MCC = 0.64. (17) Potential antihyperlipidemic activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs. Model description: Training set N = 185,
Test set N = 24, Sensitivity = 0.75, Specificity = 0.92, Accuracy = 0.83, MCC = 0.68. (18) Potential antihypertensive activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved
drugs. Model description: Training set N = 554, Test set N = 111, Sensitivity = 0.89, Specificity = 0.81, Accuracy = 0.85, MCC = 0.70. (19) Potential anti-inflammatory activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially
active compounds. Training set consists of approved drugs. Model description: Training set N = 598, Test set N = 93, Sensitivity = 0.86, Specificity = 0.84, Accuracy = 0.85, MCC = 0.69.(20) Potential activity against migraine. Cutoff is 0.5.
Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set N = 515, Test
set N = 98, Sensitivity = 0.81, Specificity = 0.84, Accuracy = 0.83, MCC = 0.65. (21) Potential antifungal activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs. Model
description: Training set N = 172, Test set N = 47, Sensitivity = 0.90, Specificity = 0.88, Accuracy = 0.89, MCC = 0.79. (22) Potential activity against obesity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds.
Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set N = 472, Test set N = 75, Sensitivity = 0.89, Specificity = 0.97, Accuracy = 0.93,
MCC = 0.87. (23) Potential anti-osteoporosis activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds
with in vivo activity. Model description: Training set N = 595, Test set N = 86, Sensitivity = 0.84, Specificity = 0.85, Accuracy = 0.85, MCC = 0.70. (24) Potential analgesic activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially
active compounds. Training set consists of approved drugs. Model description: Training set N = 525, Test set N = 84, Sensitivity = 0.92, Specificity = 0.67, Accuracy = 0.79, MCC = 0.60.(25) Potential activity against Parkinson’s disease.
Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training
set N = 298, Test set N = 49, Sensitivity = 0.96, Specificity = 0.96, Accuracy = 0.96, MCC = 0.92.(26) Potential activity against psoriasis. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists
of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set N = 199, Test set N = 32, Sensitivity = 0.93, Specificity = 0.82, Accuracy = 0.89, MCC = 0.74. (27)
Potential activity against schizophrenia. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo
activity. Model description: Training set N = 616, Test set N = 93, Sensitivity = 0.89, Specificity = 0.91, Accuracy = 0.90, MCC = 0.80.(28) Potential activity against skin diseases. Cutoff is 0.5. Values higher than 0.5 indicate potentially
active compounds. Training set consists of approved drugs. Model description: Training set N = 255, Test set N = 36, Sensitivity = 1.00, Specificity = 0.76, Accuracy = 0.86, MCC = 0.76. (29) Potential antithrombotic activity. Cutoff
is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved drugs, drug candidates in clinical trials and preclinical compounds with in vivo activity. Model description: Training set
N = 453, Test set N = 80, Sensitivity = 0.98, Specificity = 0.95, Accuracy = 0.97, MCC = 0.93.(30) Potential antiviral activity. Cutoff is 0.5. Values higher than 0.5 indicate potentially active compounds. Training set consists of approved
drugs. Model description: Training set N = 206, Test set N = 35, Sensitivity = 0.92, Specificity = 0.95, Accuracy = 0.94, MCC = 0.88. (31) Potential to be mutagenic (AMES positive), range from 0 to 1. A value of 1 is AMES positive
(mutagenic), and a value of 0 is AMES negative (non-mutagenic). Cutoff is 0.5. Values close to zero are preferable. The AMES assay is based upon the reversion of mutations in the histidine operon in the bacterium Salmonella
enterica sv Typhimurium. (32) Potential for causing anemia. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing anemia in vivo. Model organisms: human.
Model description: Training set N = 324, Test set N = 51, Sensitivity = 0.82, Specificity = 0.90, Accuracy = 0.86, MCC = 0.72. (33) Potential for inducing carcinogenicity in rats and mice. Cutoff is 0.5. Values higher than 0.5 indicate
potentially toxic compounds. Training set consists of chemicals and drugs causing carcinogenicity in vivo. Model organisms: mouse, rat. Model description: Training set N = 1210, Test set N = 185, Sensitivity = 0.96, Specificity = 0.90,
Accuracy = 0.93, MCC = 0.86. (34) Potential for inducing carcinogenicity in female mice. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing carcinogenicity
in vivo. Model organisms: female mice. Model description: Training set N = 640, Test set N = 94, Sensitivity = 0.90, Specificity = 0.93, Accuracy = 0.92, MCC = 0.83. (35) Potential for inducing carcinogenicity in male mice. Cutoff is 0.5.
Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing carcinogenicity in vivo. Model organisms: mouse male. Model description: Training set N = 584, Test set
N = 93, Sensitivity = 0.91, Specificity = 0.88, Accuracy = 0.89, MCC = 0.78. (36) Potential for inducing carcinogenicity in female rats. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of
chemicals and drugs causing carcinogenicity in vivo. Model organisms: female rat. Model description: Training set N = 667, Test set N = 120, Sensitivity = 0.90, Specificity = 0.96, Accuracy = 0.93, MCC = 0.86. (37) Potential for inducing
carcinogenicity in male rats. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing carcinogenicity in vivo. Model organisms: male rat. Model description:
Training set N = 715, Test set N = 117, Sensitivity = 0.92, Specificity = 0.88, Accuracy = 0.90, MCC = 0.79. (38) Potential for inducing cardiotoxicity. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training
set consists of chemicals and drugs causing cardiotoxicity in vivo. Model organisms: mouse, rat, human. Model description: Training set N = 143, Test set N = 30, Sensitivity = 0.80, Specificity = 1.00, Accuracy = 0.90, MCC = 0.82. (39)
Growth inhibition of MCF7 cell line (human caucasian breast adenocarcinoma), pGI50. Cutoff is 6. Values from 6 to 8 correspond to a toxic metabolite, values less than 6 are preferable, values less than 3 are more preferable and less
toxic. Model description: N = 1474, R2 = 0.9, RMSE = 0.05. (40) Potential for inducing epididymis toxicity. Training set consists of chemicals and drugs causing epididymis toxicity in vivo. Model organisms: mouse, rat, human. Cutoff
is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Model description: Training set N = 252, Test set N = 42, Sensitivity = 0.90, Specificity = 0.86, Accuracy = 0.88, MCC = 0.76. (41) Potential for inducing genotoxicity.
Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing genotoxicity in vivo. Model organisms: mouse, rat. Model description: Training set N = 372, Test
set N = 86, Sensitivity = 0.75, Specificity = 0.84, Accuracy = 0.79, MCC = 0.59. (42) Potential for inducing hepatotoxicity. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals
and drugs causing hepatotoxicity in vivo. Model organisms: mouse, rat, human. Model description: Training set N = 1380, Test set N = 231, Sensitivity = 0.73, Specificity = 0.88, Accuracy = 0.81, MCC = 0.62. (43) Potential for inducing
kidney necrosis. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing renal necrosis in vivo. Model organisms: mouse, rat, human. Model description:
Training set N = 221, Test set N = 42, Sensitivity = 0.96, Specificity = 1.00, Accuracy = 0.98, MCC = 0.95. (44) Potential for inducing kidney weight gain. Cutoff is 0.5. The values higher than 0.5 indicate potentially toxic compounds.
Training set consists of chemicals and drugs causing kidney weight gain in vivo. Model organisms: mouse, rat. Model description: Training set N = 240, Test set N = 49, Sensitivity = 0.95, Specificity = 1.00, Accuracy = 0.98, MCC = 0.96.

B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117

(45)
Potential for inducing liver cholestasis. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing cholestasis in vivo. Model organisms: mouse, rat, human.
Model description: Training set N = 218, Test set N = 35, Sensitivity = 0.79, Specificity = 0.67, Accuracy = 0.74, MCC = 0.46. (46) Potential for inducing liver lipid accumulation. Cutoff is 0.5. Values higher than 0.5 indicate potentially
toxic compounds. Training set consists of chemicals and drugs causing lipid accumulation in vivo. Model organisms: mouse, rat, human. Model description: Training set N = 172, Test set N = 28, Sensitivity = 0.80, Specificity = 0.85,
Accuracy = 0.82, MCC = 0.64. (47) Potential for inducing liver necrosis. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing hepatic necrosis in vivo. Model
organisms: mouse, rat, human. Model description: Training set N = 300, Test set N = 57, Sensitivity = 0.91, Specificity = 0.91, Accuracy = 0.91, MCC = 0.82. (48) Potential for inducing liver weight gain. Cutoff is 0.5. Values higher
than 0.5 indicate potential liver weight-changing compounds. Training set consists of chemicals and drugs causing liver weight gain in vivo. Model organisms: mouse, rat. Model description: Training set N = 292, Test set N = 52,
Sensitivity = 1.00, Specificity = 1.00, Accuracy = 1.00, MCC = 1.00. (49) Maximum Recommended Therapeutic Dose, log mg/kg-bm/day, range is from −5 to 3. Cutoff is 0.5. Chemicals with high log MRTDs can be classified as midly
toxic compounds, chemicals with low log MRTDs as highly toxic compounds. Model description: N = 1209, R2 = 0.86, RMSE = 0.42. (50) Potential for causing nasal pathology. Training set consists of chemicals and drugs causing
nasal pathology in vivo. Model organisms: mouse, rat, human. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Model description: Training set N = 246, Test set N = 47, Sensitivity = 1.00, Specificity = 0.93,
Accuracy = 0.96, MCC = 0.92. (51) Potential for inducing nephron injury. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Training set consists of chemicals and drugs causing nephron injury in vivo. Model
organisms: mouse, rat, human. Model description: Training set N = 598, Test set N = 109, Sensitivity = 0.91, Specificity = 1.00, Accuracy = 0.96, MCC = 0.93. (52) Potential for inducing nephrotoxicity. Cutoff is 0.5. Values higher than 0.5
indicate potentially toxic compounds. Training set consists of chemicals and drugs causing nephrotoxicity in vivo. Model organisms: mouse, rat, human. Model description: Training set N = 847, Test set N = 154, Sensitivity = 0.90,
Specificity = 0.84, Accuracy = 0.87, MCC = 0.74. (53) Potential for inducing neurotoxicity. Training set consists of chemicals and drugs causing neurotoxicity in vivo. Model organisms: mouse, rat, human. Cutoff is 0.5. Values higher
than 0.5 indicate potentially toxic compounds. Model description: Training set N = 175, Test set N = 34, Sensitivity = 0.94, Specificity = 0.94, Accuracy = 0.94, MCC = 0.88. (54) Potential for inducing pulmonary toxicity. Training set
consists of chemicals and drugs causing pulmonary toxicity in vivo. Model organisms: mouse, rat, human. Cutoff is 0.5. Values higher than 0.5 indicate potentially toxic compounds. Model description: Training set N = 482,
Test set N = 87, Sensitivity = 0.89, Specificity = 0.88, Accuracy = 0.89, MCC = 0.77. (55) Skin sensitization potential expressed as effective concentration 3, EC3%. Values higher than 10 indicate weak and moderate sensitizers. Model
description: N = 89, R2 = 0.67, RMSE = 22.56. (56) Potential for inducing testicular toxicity. Training set consists of chemicals and drugs causing testicular toxicity in vivo. Model organisms: mouse, rat, human. Cutoff is 0.5. Values
higher than 0.5 indicate potentially toxic compounds. Model description: Training set N = 439, Test set N = 88, Sensitivity = 0.81, Specificity = 0.85, Accuracy = 0.83, MCC = 0.66.

115
116 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117
4. Conclusions
The development of mathematical models using results from
experimental studies on multiple length scales assists for better
understanding of massive challenges in different scientific disci-
plines [59–62]. Such as in this work, several molecular screening
methods were performed and it is found that oxazolone and imida-
zolone derivatives reveal similar/better interaction energy profiles
compared to the FDA approved sartan molecules at the binding
site of the monomer and homodimer AT1 receptor structures. In
addition, virtual screening of around 18 million small drug-like
compounds from ZINC database were screened and their docking
scores were compared with known AT1 antagonists. MD simu-
lations and post-processing MD analyses were implemented for
hERG-neutral hit compounds which were derived from de novo
design and virtual screening studies. Results of this study can be
useful for designing of novel, safe and selective AT1 antagonists.
Acknowledgements
The numerical calculations reported in this paper were par-
tially performed at TUBITAK ULAKBIM, High Performance and Grid
Computing Center (TRUBA resources). This work was financially
supported by the Scientific and Technological Research Council of
Turkey (TUBITAK), Project No: 214Z122.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.jmgm.2017.10.011.
References
[1] WHO, A Global Brief on Hypertension: Silent Killer, Global Public Health
Crisis, 2013.
[2] P. Naik, P. Murumkar, R. Giridhar, M.R. Yadav, Angiotensin II receptor type 1
(AT 1) selective nonpeptidic antagonists—A perspective, Bioorg. Med. Chem.
18 (2010) 8418–8456.
[3] R.M. Carey, AT1 receptors, angiotensin receptor blockade, and clinical
hypertensive disease, in: Renin Angiotensin System and Cardiovascular
Disease, Springer, 2009, 2017, pp. 59–79.
[4] R.M. Carey, Pathophysiology of primary hypertension, Compr. Physiol. (2008).
[5] M. De Gasparo, K. Catt, T. Inagami, J. Wright, T. Unger, International union of
pharmacology: XXIII. The angiotensin II receptors, Pharmacol. Rev. 52 (2000)
415–472.
[6] G. Agelis, P. Roumelioti, A. Resvani, S. Durdagi, M.-E. Androutsou, K.
Kelaidonis, et al., An efficient synthesis of a rationally designed 1, 5
disubstituted imidazole AT1 angiotensin II receptor antagonist: reorientation
of imidazole pharmacophore groups in losartan reserves high receptor affinity
and confirms docking studies, J. Comput. Aided Mol. Des. 24 (2010) 749–758.
[7] T. Dolley-Hitze, F. Jouan, B. Martin, S. Mottier, J. Edeline, O. Moranne, et al.,
Angiotensin-2 receptors (AT1-R and AT2-R), new prognostic factors for renal
clear-cell carcinoma, Br. J. Cancer 103 (2010) 1698–1705.
[8] J. Ge, N.M. Barnes, Alterations in angiotensin AT1 and AT2 receptor subtype
levels in brain regions from patients with neurodegenerative disorders, Eur. J.
Pharmacol. 297 (1996) 299–306.
[9] R.H. Bradbury, C.P. Allott, M. Dennis, E. Fisher, J.S. Major, B.B. Masek, et al.,
New nonpeptide angiotensin II receptor antagonists 2. Synthesis, biological
properties, and structure-activity relationships of
2-alkyl-4-(biphenylylmethoxy) quinoline derivatives, J. Med. Chem. 35 (1992)
4027–4038.
[10] H.R. Brunner, H. Gavras, Angiotensin blockade for hypertension: a promise
fulfilled, Lancet 359 (2002) 990–992.
[11] A. Cappelli, C. Nannicini, A. Gallelli, G. Giuliani, S. Valenti, G.l.P. Mohr, et al.,
Design, synthesis, and biological evaluation of AT1 angiotensin II receptor
antagonists based on the pyrazolo [3, 4-b] pyridine and related
heteroaromatic bicyclic systems, J. Med. Chem. 51 (2008) 2137–2146.
[12] S.E. Easthope, B. Jarvis, Candesartan cilexetil, Drugs 62 (2002) 1253–1287.
[13] M.A. Ismail, S. Barker, D.A. Abou El Ella, K.A. Abouzid, R.A. Toubar, M.H. Todd,
Design and synthesis of new tetrazolyl-and
carboxy-biphenylylmethyl-quinazolin-4-one derivatives as angiotensin II AT1
receptor antagonists, J. Med. Chem. 49 (2006) 1526–1535.
[14] M.P. Maillard, J. Rossat, H.-R. Brunner, M. Burnier, Tasosartan, enoltasosartan,
and angiotensin II receptor blockade: the confounding role of protein binding,
J. Pharmacol. Exp. Ther. 295 (2000) 649–654.
[15] B.B. Masek, A. Merchant, J.B. Matthew, Molecular shape comparison of
angiotensin II receptor antagonists, J. Med. Chem. 36 (1993) 1230–1238.
[16] T. Kellici, A. Tzakos, T. Mavromoustakos, Rational drug design and synthesis of
molecules targeting the angiotensin II type 1 and type 2 receptors, Molecules
20 (2015) 3868.
[17] F.K. Tahsin, N. Dimitrios, K. Eftichia, Z. Maria, Z. Panagiotis, P. Constantinos,
et al., Leveraging NMR and X-ray data of the free ligands to build better drugs
targeting angiotensin II type 1 G-Protein coupled receptor, Curr. Med. Chem.
23 (2016) 36–59.
[18] S. Durdagi, S. Deshpande, H.J. Duff, S.Y. Noskov, Modeling of open, closed, and
open-inactivated states of the hERG1 channel: structural mechanisms of the
state-dependent drug binding, J. Chem. Inf. Model. 52 (2012) 2760–2774.
[19] S. Durdagi, H.J. Duff, S.Y. Noskov, Combined receptor and ligand-based
approach to the universal pharmacophore model development for studies of
drug blockade to the hERG1 pore domain, J. Chem. Inf. Model. 51 (2011)
463–474.
[20] S. Durdagi, J. Guo, J.P. Lees-Miller, S.Y. Noskov, H.J. Duff, Structure-guided
topographic mapping and mutagenesis to elucidate binding sites for the
human ether-a-go-go-related gene 1 potassium channel (KCNH2) activator
NS1643, J. Pharmacol. Exp. Ther. 342 (2012) 441–452.
[21] S. Durdagi, J. Subbotina, J. Lees-Miller, J.J. Guo, H.Y. Duff, S. Noskov, Insights
into the molecular mechanism of hERG1 channel activation and blockade by
drugs, Curr. Med. Chem. 17 (2010) 3514–3532.
[22] J. Subbotina, V. Yarov-Yarovoy, J. Lees-Miller, S. Durdagi, J. Guo, H.J. Duff, et al.,
Structural refinement of the hERG1 pore and voltage-sensing domains with
ROSETTA-membrane and molecular dynamics simulations, Proteins Struct.
Funct. Bioinf. 78 (2010) 2922–2934.
[23] W. Wang, R. MacKinnon, Cryo-EM structure of the open human
ether-a-go-go-related K+ channel hERG, Cell 169 (2017) 422–430 (e10).
[24] H. Zhang, H. Unal, C. Gati, G.W. Han, W. Liu, N.A. Zatsepin, et al., Structure of
the angiotensin receptor revealed by serial femtosecond crystallography, Cell
161 (2015) 833–844.
[25] T. Tuccinardi, V. Calderone, S. Rapposelli, A. Martinelli, Proposal of a new
binding orientation for non-peptide AT1 antagonists: homology modeling,
docking and three-dimensional quantitative structure-activity relationship
analysis, J. Med. Chem. 49 (2006) 4305–4316.
[26] T. Okada, M. Sugihara, A.-N. Bondar, M. Elstner, P. Entel, V. Buss, The retinal
conformation and its environment in rhodopsin in light of a new 2.2 Å crystal
structure, J. Mol. Biol. 342 (2004) 571–583.
[27] H. Zhang, H. Unal, R. Desnoyer, G.W. Han, N. Patel, V. Katritch, et al., Structural
basis for ligand recognition and functional selectivity at angiotensin receptor,
J. Biol. Chem. 290 (2015) 29127–29139.
[28] T.F. Kellici, D. Ntountaniotis, G. Liapakis, A.G. Tzakos, T. Mavromoustakos, The
dynamic properties of angiotensin II type 1 receptor inverse agonists in
solution and in the receptor site, Arab. J. Chem. (2016), http://dx.doi.org/10.
1016/j.arabjc.2016.11.014.
[29] G.M. Sastry, M. Adzhigirey, T. Day, R. Annabhimoju, W. Sherman, Protein and
ligand preparation: parameters, protocols, and influence on virtual screening
enrichments, J. Comput. Aided Mol. Des. 27 (2013) 221–234.
[30] D.C. Bas, D.M. Rogers, J.H. Jensen, Very fast prediction and rationalization of
pKa values for protein–ligand complexes, Proteins Struct. Funct. Bioinf. 73
(2008) 765–783.
[31] H. Li, A.D. Robertson, J.H. Jensen, Very fast empirical prediction and
rationalization of protein pKa values, Proteins Struct. Funct. Bioinf. 61 (2005)
704–721.
[32] J.L. Banks, H.S. Beard, Y. Cao, A.E. Cho, W. Damm, R. Farid, et al., Integrated
modeling program, applied chemical theory (IMPACT), J. Comput. Chem. 26
(2005) 1752–1780.
[33] S.W. Siu, K. Pluhackova, R.A. Böckmann, Optimization of the OPLS-AA force
field for long hydrocarbons, J. Chem. Theory Comput. 8 (2012) 1459–1470.
[34] J.J. Irwin, B.K. Shoichet, ZINC-a free database of commercially available
compounds for virtual screening, J. Chem. Inf. Model. 45 (2005) 177–182.
[35] J.R. Greenwood, D. Calkins, A.P. Sullivan, J.C. Shelley, Towards the
comprehensive, rapid, and accurate prediction of the favorable tautomeric
states of drug-like molecules in aqueous solution, J. Comput. Aided Mol. Des.
24 (2010) 591–604.
[36] J.C. Shelley, A. Cholleti, L.L. Frye, J.R. Greenwood, M.R. Timlin, M. Uchimaya,
Epik: a software program for pK a prediction and protonation state generation
for drug-like molecules, J. Comput. Aided Mol. Des. 21 (2007) 681–691.
[37] R.A. Friesner, J.L. Banks, R.B. Murphy, T.A. Halgren, J.J. Klicic, D.T. Mainz, et al.,
Glide: a new approach for rapid, accurate docking and scoring: 1. Method and
assessment of docking accuracy, J. Med. Chem. 47 (2004) 1739–1749.
[38] R.A. Friesner, R.B. Murphy, M.P. Repasky, L.L. Frye, J.R. Greenwood, T.A.
Halgren, et al., Extra precision glide: docking and scoring incorporating a
model of hydrophobic enclosure for protein-ligand complexes, J. Med. Chem.
49 (2006) 6177–6196.
[39] T.A. Halgren, R.B. Murphy, R.A. Friesner, H.S. Beard, L.L. Frye, W.T. Pollard,
et al., Glide: a new approach for rapid, accurate docking and scoring: 2.
Enrichment factors in database screening, J. Med. Chem. 47 (2004)
1750–1759.
[40] M.L. Verdonk, J.C. Cole, M.J. Hartshorn, C.W. Murray, R.D. Taylor, Improved
protein–ligand docking using GOLD, Proteins Struct. Funct. Bioinf. 52 (2003)
609–623.
[41] H. Wu, D. Wacker, M. Mileni, V. Katritch, G.W. Han, E. Vardy, et al., Structure
of the human [kgr]-opioid receptor in complex with JDTic, Nature 485 (2012)
327–332.
 B. Aksoydan et al. / Journal of Molecular Graphics and Modelling 79 (2018) 103–117
[42] S. Jo, T. Kim, W. Im, Automated builder and database of protein/membrane
complexes for molecular dynamics simulations, PLoS One 2 (2007) e880.
[43] J. Lee, X. Cheng, J.M. Swails, M.S. Yeom, P.K. Eastman, J.A. Lemkul, et al.,
CHARMM-GUI input generator for NAMD, GROMACS, AMBER, OpenMM, and
CHARMM/OpenMM simulations using the CHARMM36 additive force field, J.
Chem. Theory Comput. 12 (2015) 405–413.
[44] E.L. Wu, X. Cheng, S. Jo, H. Rui, K.C. Song, E.M. Dávila-Contreras, et al.,
CHARMM-GUI Membrane Builder toward realistic biological membrane
simulations, J. Comput. Chem. 35 (2014) 1997–2004.
[45] M.A. Lomize, A.L. Lomize, I.D. Pogozheva, H.I. Mosberg, OPM: orientations of
proteins in membranes database, Bioinformatics 22 (2006) 623–625.
[46] A.L. Lomize, I.D. Pogozheva, M.A. Lomize, H.I. Mosberg, Positioning of proteins
in membranes: a computational approach, Protein Sci. 15 (2006) 1318–1333.
[47] B. Hess, C. Kutzner, D. Van Der Spoel, E. Lindahl, GROMACS 4: algorithms for
highly efficient, load-balanced, and scalable molecular simulation, J. Chem.
Theory Comput. 4 (2008) 435–447.
[48] D. van der Spoel, E. L, B. Hess, GROMACS Development Team, GROMACS
Version 4.6.5, 2013.
[49] H.J. Berendsen, D. van der Spoel, R. van Drunen, GROMACS: a
message-passing parallel molecular dynamics implementation, Comput.
Phys. Commun. 91 (1995) 43–56.
[50] H.J. Berendsen, J.v. Postma, W.F. van Gunsteren, A. DiNola, J. Haak, Molecular
dynamics with coupling to an external bath, J. Chem. Phys. 81 (1984)
3684–3690.
[51] T. Darden, D. York, L. Pedersen, Particle mesh Ewald: an N·log (N) method for
Ewald sums in large systems, J. Chem. Phys. 98 (1993) 10089–10092.
[52] W.G. Hoover, Canonical dynamics: equilibrium phase-space distributions,
Phys. Rev. A 31 (1985) 1695.
[53] G. Rastelli, A. Del Rio, G. Degliesposti, M. Sgobba, Fast and accurate
predictions of binding free energies using MM-PBSA and MM-GBSA, J.
Comput. Chem. 31 (2010) 797–810.
117
[54] G. Milligan, G protein-coupled receptor dimerization: function and ligand
pharmacology, Mol. Pharmacol. 66 (2004) 1–7.
[55] S. Terrillon, M. Bouvier, Roles of G-protein-coupled receptor dimerization,
EMBO Rep. 5 (2004) 30–34.
[56] B.M. Young, E. Nguyen, M.A. Chedrawe, J.K. Rainey, D.J. Dupre, Differential
contribution of transmembrane domains IV, V, VI, and VII to human
angiotensin II type 1 receptor homomer formation, J. Biol. Chem. 292 (2017)
3341–3350.
[57] S. Durdagi, R.E. Salmas, M. Stein, M. Yurtsever, P. Seeman, Binding interactions
of dopamine and apomorphine in D2High and D2Low states of human
dopamine D2 receptor using computational and experimental techniques,
ACS Chem. Neurosci. 7 (2016) 185–195.
[58] D. Tu, Y. Liao, A. Zou, Y. Du, X. Wang, L. Li, Electropharmacological properties
of telmisartan in blocking hKv1: 5 and HERG potassium chanenls expressed
on Xenopus laevis oocytes, Acta Pharmacol. Sin. 29 (2008) 913–922.
[59] M. Zervou, Z. Cournia, C. Potamitis, G. Patargias, S. Durdagi, S.G. Grdadolnik, T.
Mavromoustakos, Insights into the molecular basis of action of AT1
antagonist losartan using a combined NMR spectroscopy and computational
approach, BBA Biomembr. 2838 (2014) 1031–1046.
[60] E. Kritsi, C. Potamitis, S. Durdagi, P. Zoumpoulakis, S.G. Grdadolnik, T.
Mavromoustakos, Molecular Insights into the AT1 antagonism based on
biophysical and in silico studies of telmisartan, Med. Chem. Res. 22 (2013)
4842–4857.
[61] T. Mavromoustakos, G. Agelis, S. Durdagi, AT1 antagonists: a patent reviewe
(2008–2012), Expert Opin. Ther. Pat. 23 (2013) 1483–1494.
[62] G. Agelis, A. Resvani, S. Durdagi, S. Spyridaki, T. Tumpva, J. Slaninova, P.
Giannopoulos, D. Vlahakos, G. Liapakis, T. Mavromoustakos, J. Matsoukas, The
discovery of new potent non-peptide Angiotensin II AT1 receptor blockers: a
concise synthesis, molecular docking studies and biological evaluation of
N-substituted 5-buthylimidazole derivatives, Eur. J. Med. Chem. 55 (2012)
358–374.