Virtual screening of National Cancer Institute database for claudin-4 inhibitors: Synthesis, biological evaluation, and molecular dynamics studies

Article abstract of DOI:10.1002/jcb.28147

Claudin-4 (CLDN4) is a vital member of tight-junction proteins that is often overexpressed in cancer and other malignancies. The three-dimensional structure of human CLDN4 was constructed based on homology modeling approach. A total of 265 242 molecules from the National Cancer Institute (NCI) database has been utilized as a dataset for this study. In the present work, structure-based virtual screening is performed with the NCI database using Glide. By molecular docking, 10 candidate molecules with high scoring functions, which binds to the active site of CLDN4 were identified. Subsequently, molecular dynamics simulations of membrane protein were used for optimization of the top-three lead compounds (NCI110039, NCI344682, and NCI661251) with CLDN4 in a dynamic system. The lead molecule from NCI database NCI11039 (purpurogallin carboxylic acid) was synthesized and cytotoxic properties were evaluated with A549, MCF7 cell lines. Our docking and dynamics simulations predicted that ARG31, ASN142, ASP146, and ARG158 as critically important residues involved in the CLDN4 activity. Finally, three lead candidates from the NCI database were identified as potent CLDN4 inhibitors. Cytotoxicity assays had proved that purpurogallin carboxylic acid had an inhibitory effect towards breast (MCF7) and lung (A549) cancer cell lines. Computational insights and in vitro (cytotoxicity) studies reported in this study are expected to be helpful for the development of novel anticancer agents.

Full text of DOI:10.1002/jcb.28147

Received: 2 August 2018
DOI: 10.1002/jcb.28147
Accepted: 5 November 2018
|
R E S E A R C H A R T I C L E
Virtual screening of National Cancer Institute database for
claudin‐4 inhibitors: Synthesis, biological evaluation, and
molecular dynamics studies
Majji Rambabu
|
Sivaraman Jayanthi
Department of Biotechnology,
Abstract
Computational Drug Design Lab, School
of Bio Sciences and Technology, Vellore
Institute of Technology, Vellore, India
Claudin‐4 (CLDN4) is a vital member of tight‐junction proteins that is often
overexpressed in cancer and other malignancies. The three‐dimensional structure
of human CLDN4 was constructed based on homology modeling approach. A total
of 265 242 molecules from the National Cancer Institute (NCI) database has been
utilized as a dataset for this study. In the present work, structure‐based virtual
screening is performed with the NCI database using Glide. By molecular docking,
Correspondence
Sivaraman Jayanthi, Computational Drug
Design Lab, School of Bio Sciences and
Technology, Vellore Institute of
Technology, Vellore 632014, India.
Email: jayanthi.s@vit.ac.in
10 candidate molecules with high scoring functions, which binds to the active site
of CLDN4 were identified. Subsequently, molecular dynamics simulations of
Funding information
DST‐SERB, India., Grant/Award Number:
SB/YS/LS‐299/2013
membrane protein were used for optimization of the top‐three lead compounds
(NCI110039, NCI344682, and NCI661251) with CLDN4 in a dynamic system. The
lead molecule from NCI database NCI11039 (purpurogallin carboxylic acid) was
synthesized and cytotoxic properties were evaluated with A549, MCF7 cell lines.
Our docking and dynamics simulations predicted that ARG31, ASN142, ASP146,
and ARG158 as critically important residues involved in the CLDN4 activity.
Finally, three lead candidates from the NCI database were identified as potent
CLDN4 inhibitors. Cytotoxicity assays had proved that purpurogallin carboxylic
acid had an inhibitory effect towards breast (MCF7) and lung (A549) cancer cell
lines. Computational insights and in vitro (cytotoxicity) studies reported in this
study are expected to be helpful for the development of novel anticancer agents.
K E Y W O R D S
claudin‐4, homology modeling, molecular dynamics, virtual screening
1
|
INTRODUCTION
Usually, tumor development is associated with disruption
of TJs and downregulation of CLDNs. They have been
Claudins (CLDNs) were found to be vital and adequate to
accounted for few cancers and it is related to a poor
5-7
form tight‐junction (TJ) strands and accounts for some of
prognosis or metastatic sickness. In comparison to other
strong tumors, the peritoneal spread is the most widely
known technique for ovarian malignancy. Tumor cells
subside the ovary into the peritoneal liquid where it
scatters all over the abdominal cavity. Thus, it attaches to
the mesothelial cell covering which leads to metastatic
1,2
the specific variability of different barriers. CLDNs were
one of the important class of cellular adhesion molecules,
which had a key role in cell polarity, epithelial cell sheets
arrangement, and in paracellular transport. Expression of
CLDNs in normal cells is specific and its altered
3
,4
8
expression is distinguished in different cancer types.
developments. Nearly 70% of the ovarian growth tissues
|
J Cell Biochem. 2018;1-13.
wileyonlinelibrary.com/journal/jcb
© 2018 Wiley Periodicals, Inc.
1

2
RAMBABU AND JAYANTHI
|
expressed claudin‐4 (CLDN4) were distinctively expressed
database. Virtual screening and dynamics help us to find
the top‐three molecules. Subsequently, NCI110039 (purpur-
ogallin carboxylic acid) has been synthesized and in vitro
studies were conducted. The in silico studies were validated
through biological evaluation. This rational approach was
adopted for screening the potential inhibitors against
CLDN4, which would be helpful in decrypting the binding
mechanism that leads to efficient inhibitory action CLDN4.
across ovarian tumor subtypes with the minimum
9
expression in pure cell subtype. The function of CLDN4
is managed by phosphorylation via kinases as well as
forced or knockdown expression. For instance, phosphor-
10
ylation of CLDN4 by cAMP‐dependent protein kinase or
1
1
protein kinase C (PKC) increases paracellular perme-
ability in ovarian disease cells through mislocalization of
CLDNs. In ovarian cancer cells, PKCα enactment results
in the mislocalization of CLDN4 with the diminished TJ
2
2
|
MATERIALS AND METHODS
Data set preparation
12
barrier uprightness in human pancreatic cancer cells.
CLDN4 advances the creation of factors that fortify
angiogenesis in both in vitro and in vivo processes which
recommend its aggressive angiogenic part in ovarian
.1
|
2
2
The NCI database is used for this study. In Schrodinger
suite, the database is curated by visualizing the whole
database in Canvas. The improper molecules were
excluded from this criterion, such as molecular weight
(high and low) of less than 150 and more than 500,
number of aromatic rings, and structure similarity. As
per the above criteria, the database comprises of 177 236
molecules and exported into the SDF format in maestro
were used in ligand preparation.
13
cancer. CLDN4 in ovarian cancer depends on the
hypothesis that ovarian malignancies start from typical
ovarian surface epithelial cells which do not express
CLDN4. In this manner, the overexpression of CLDN4 has
been appeared to promote the progression of the ovarian
disease. Constrained expression of CLDN4 in ovarian
epithelial cells increases invasive behavior by instigating
14
the activation of matrix metalloproteinase. The deloca-
lization of CLDN proteins from the cell membrane is
regular among the changed cells and in the ovarian
growth which connects with tumor cell migration and
2
.2
|
Ligand structure preparation
3,4
invasion.
The mechanism of the increased CLDN4
Curated NCI compounds were used in this study. Prepara-
tion of ligands for virtual screening and docking had been
done using LigPrep 3.4 module of the Schrodinger suite
2015‐2. All ligands were prepared by utilizing the force field
called OPLS‐2005. The ionization state was fixed to produce
every single conceivable state. The molecular weight was
set to be less than 150 and greater than 500. Ligand original
state using “EpiK,” tautomers were generated for the
ligands. Utilizing the stereoisomers column maintains the
same chiralities and produce 32 at most for each ligand was
allowed. Moreover, generated low‐energy ring conforma-
tions were provided one for each ligand in the output file of
ligprep.
expression in the ovarian carcinoma is thought to be the
consequence of epigenetic adjustments of the CLDN
promoter areas. These areas in the tumor cells bring
results in increased cell survival, motility, and inva-
14-16
sion.
Ovarian tumors of various subtypes like muci-
nous, serous, undifferentiated, clear cell, and endometrioid
carcinomas had been found exceptional in the CLDN4
disease state, yet normal ovarian surface epithelium does
1
3,14,17-19
not express.
ovarian growth.
CLDN4 had a prognostic part in the
Epithelial ovarian carcinoma remains
13,18
gynecologic malignancy with the most noteworthy mor-
20
tality rate. Even though the mechanism of CLDN4 is yet
2
1
to be completely elucidated.
In this investigation, we aim to identify potential
inhibitors for human CLDN4 with the National Cancer
Institute (NCI) database. The curated NCI database is
subjected to virtual screening for the CLDN4 target.
Further, high throughput virtual screening (HTVS) had
been performed to screen the ligand molecules. The docked
complexes were screened based on the postdocking analysis
which includes binding energy, hydrogen bond (H‐bond)
interactions, and binding mode. In addition, the pharma-
cological properties of these inhibitors were validated using
the absorption, distribution, metabolism, and excretion
2
.3
|
Homology modeling
Homology modeling of the CLDN4 was performed by using
the Schrodinger 10.2 software. The human CLDN4 was
retrieved from Swiss‐Prot (http://www.uniprot.org/) (acces-
sion number: O14493) consists of 209 amino acids. The
crystal structure of claudin‐15 was used as template
(CLDN15) (Mus musculus) (PDB ID: 4P79). The template
and target sequence were aligned by using ClustalW.
Subsequently, secondary structure prediction was carried
out for a target and template that helps to find the best
model. Further, homology of CLDN4 was based upon the
energy‐based method. The modeled structure of CLDN4 was
validated by PROCHECK in the SAVeS server.
23
(ADME) properties and Lipinski’s rule of five. Moreover,
the performed molecular dynamic simulations were done
for apo and CLDN4 with lead molecules from the NCI

RAMBABU AND JAYANTHI
3
|
2
.4
|
Preparation of protein
generated with a grid sizing 80 × 80 × 80 Å through
coordinates X = −24.05, Y = 1.65, Z = 2.92 and the small
molecules to be docked. It was chosen without exceeding
The modeled structure of CLDN4 is used for protein
preparation. Protein Preparation Wizard is available in the
Glide. Using import and process results in the elements
like assigned bond orders, disulfide bonds, deleted all
waters, hydrogens added, and capped termini. Under the
refinement, the H‐bond assignment enhances to alter the
sample water symmetry and possible positions of hydroxyl
hydrogen and thiol atoms, tautomers and protonation
states of His residues and Chi “flip” assignments chosen
through default script allowed in Schrodinger. The range
of pH was set to 7.0 and using the OPLS‐2005 force field
1
0 Å. In ligand docking, the protein is flexible, permits
the rotation of thiol and hydroxyl group of TYR33,
TYR67, TYR139, SER157. All the ligands were docked
into the dynamic binding site by utilizing “extra
precision” algorithm of the Glide.
2
.7
|
Free energy calculation
The Prime molecular mechanics generalized born surface
area (MM/GBSA) was performed to figure out the
binding energy. This method is utilized to anticipate
the free binding energy for the set of ligands to the
protein. Binding energy is characterized as the entirety of
binding and straining energy of the ligand. The accom-
2
4,25
the protein has been minimized.
For the grid
preparation of CLDN4, Glide package had been utilized
to signify the properties and shape of the protein by
various arrangements and that gives continuously the
more exact scoring of the ligand postures.
3
1
panying equation calculates the binding energy.
ΔGbind = ΔEMM + ΔGsolv + ΔGSA.
2.5
|
Virtual screening workflow
Numerous strategies have been available for virtual
26,27
2.8
|
ADME screening
screening.
Structure‐based virtual screening is valuable
for identifying novel small molecules which bind to
protein targets and predicts binding affinity. For docking,
the issues included in making scoring functions have been
The QikProp v4.4 program incorporated inside Schro-
dinger was utilized to acquire the ADME properties of
the NCI ligands. The program was processed with
neutralized and permitted to compute almost 35 physi-
cochemical properties for all compounds. The predicted
descriptors such as H‐bond acceptors, H‐bond donors,
molecular weight, logP (octanol/water), humoral absorp-
tion were a couple of imperative properties. It addition-
ally assesses the blood‐barrier crossing and central
nervous system activity of the compound along with
the drug‐likeness of the molecules in view of Lipinski’s
28,29
recently highlighted.
The compounds were initially
filtered based on the Lipinski’s rule of five that sets the
30
criteria for a drug‐like properties. The source of the
ligands submitted as ligprep output file, combined all
input files and redistributed for sub jobs and set to identify
unique compounds by generating a unique property for
each input compound. In the filtering tab, QikProp
(ADME properties) and prefilter by the Lipinski’s rule
3
2
33,34
were performed. In the preparation tab all parameters
were set as default such as prepare ligands, target pH,
remove high‐energy tautomer states, specified 2D proper-
ties, and specified as unspecified stereocenters retain up to
four stereoisomers, generate low‐energy ring conforma-
tions one per ligand. Under receptor provided the grid of
the protein. In the docking tab, Epik state penalties for
docking were used to write interaction score for residue
within 12 Å of grid center. The scaling factor is 0.80,
partial charge cutoff 0.15 and performed Glide standard
precision (SP) for all the ligands.
rule of five and Jorgensen rule of three.
These two
had ended up as being solid standards to determine the
drug‐likeness of a synthetic compound with a specific
biological activity or pharmacological properties, that
would make it feasible orally active for a drug.
2
.9
|
MD simulations for membrane
protein
35-37
The Desmond v4.2 module
was utilized to ponder the
thermodynamic stability of the protein in apo form and
protein‐ligand complex. Set up a lipid membrane dipalmi-
toylphosphatidylcholine for protein to cover the four
transmembrane domains of the protein. TIP3P water
2.6
|
Ligand docking
38
Grid Based Ligand Docking with Energetics (Glide) v6.7
Small‐Molecule Drug Discovery Suite, 2015‐2) package
model was utilized to simulate water molecules by
utilizing an OPLS‐2005 force field. Orthorhombic periodic
boundary conditions were fixed up to indicate the shape
and size of repeating unit buffered at 10 Å distances. To
neutralize the system electrically, the appropriate three
(
in Schrodinger is utilized for ligand dockings with an
OPLS‐2005 force field. When it is confirmed that ligands
and protein are right structures to dock, the receptor was

4
RAMBABU AND JAYANTHI
|
chloride counterions were added to maintain the system
charge and place haphazard in the solvated system.
Moreover, the solvated system is minimized and relaxed
by certain OPLS‐2005 force field. The default Desmond
minimization and equilibration procedure were followed.
The dynamic simulations were accomplished with the
periodic conditions to ensemble the amount of substance,
volume and temperature (NPT). The procedure was
succeeded by running 50‐nanoseconds NPT production
simulation. Thus, obtained every 5‐picoseconds intervals
and saved the configurations. SHAKE has been applied to
all hydrogen atoms and bond lengths using default
settings. The whole Desmond simulations were performed
on the Intel (R) Xeon (R) CPU E5–2650v core processor
CPU @ 2.60GHz with 16GB RAM. Schrodinger software
was incorporated and runs under Ubuntu Linux 14.0
operating system.
the analysis, the data has been plotted using the Topspin
software.
2
.12
|
Analysis of cell viability
Cancer cells were maintained in Dulbecco’s modified
Eagle’s medium, added with 10% fetal bovine serum. The
cells were incubated in a humidified incubator at 37°C
39
with 5% CO for growth. CLDN4 has been involved in
2
many cancers including breast and lung. We have chosen
two major cell lines MCF7 (breast cancer) A549 (lung
cancer) to evaluate the inhibitor (purpurogallin car-
boxylic acid) efficiency. Cell viability of A549 and MCF7
after purpurogallin carboxylic acid treatment was esti-
mated by the 3‐[4,5‐dimethylthiazol‐2‐yl]‐2,5‐diphenyl‐
tetrazolium bromide (MTT) assay. Briefly, A549, MCF7
4
cells were seeded in a 96‐well plate at a density of 1 × 10
cells/well. Purpurogallin carboxylic acid is added on next
day by replacing the medium with fresh medium with
various concentration. After 24 hours of incubation,
2
.10
|
Synthesis of purpurogallin
carboxylic acid
1
0 µL of MTT (10 mg/mL) was added into in each well.
40
NCI110039 compound chemical name is purpurogallin
carboxylic acid and both the compounds structure remain
same. Aqueous solutions of pyrogallol (10 mmol, 1.26
in 10 mL of water) and potassium iodate (10 mmol, 2.14 g
in 10 mL of water) were added simultaneously with stirring
and ice‐cooling to the slurry of sodium gallate obtained by
adding gallic acid (12 mmol, 2.04 g) to a solution of sodium
hydrogen carbonate (12 mmol, 1.04 g) in water (20 mL) so
that addition of pyrogallol and potassium iodate mixture
was added slowly in 45 minutes. After the reaction mixture
was stirred for 30 minutes. The reaction mixture was
acidified by adding 10 N‐hydrochloric acid (15 mL),
and the reaction mixture set aside overnight. After
completion of the reaction, the mixture was filtered and
the filtrate was extracted three times with chloroform. All
the extracts of the organic layer were combined, the solvent
was removed by using vacuum distillation until precipitate
After keeping for another 4 hours in 37°C, the super-
natant is removed and 100 µL of DMSO is added in each
well. Microplate reader at a wavelength of 570 nm is used
to read the absorbance of each well. The control group
which had untreated cells was measured as 100% viability
of cells. Results are concluded as a percentage of viable
cells when related with the control group. The mitochon-
drial‐dependent reduction of MTT to formazan used to
4
2
measure cell respiration as an indicator of cell viability.
3
|
RESULTS AND DISCUSSION
CLDN4 protein has four transmembrane domains, namely
TM1, TM2, TM3, and TM4, two extracellular loops and one
cytosolic loop. Extracellular loops are named as ECL1 and
43
ECL2. ECL1 has more residues compared with ECL2
(Figure 1).
(5 g). The crude product was recrystallized with tetrahy-
drofuran/ethanol a small orange microcrystal has been
formed at melting above 320°. The molecular formula of the
synthesized compound is C H O and molecular weight is
3
.1
|
Validation of the modeled
12 8 7
structure
41
264.0270 (Supporting Information Figure S1).
A few strategies to assess the geometrical and structural
consistency of the homology model was carried out and
the similarity between target and template was found to
be 45%. Using PROCHECK, the physicochemical and
structural elements of the CLDN4 was validated. The
Ramachandran plot reveals that CLDN4 has a splendid
geometrical consistency. In the Ramachandran plot,
residues for the favorable region inside is 91.2%, allowed
region is 6.6%, generously allowed region 1.6%, and in the
disallowed region only one residue (LEU176) is found.
2
.11
|
Nuclear magnetic resonance
analysis
Nuclear magnetic resonance (NMR) spectroscopy is a
common technique used to find elements of the chemical
components particularly proton, carbon, phosphorous,
and sulfur. In this study, proton and carbon NMR was
used to confirm the compound purpurogallin carboxylic
acid using dimethyl sulfoxide (DMSO) as a solvent. After

RAMBABU AND JAYANTHI
5
|
than four binding energies had been taken as a separate
dataset to perform SP and XP dock.
3
.3
|
Molecular docking
Molecular docking was implemented to find the depth of
molecular interactions to calculate the binding energy
between ligand compounds and the CLDN4. The structure
of CLDN4 and the best compounds of SP were docked into
the active site of CLDN4. The results of docking with
compounds NCI110039, NCI344682, and NCI661251 are
shown in Table 1. The docking findings confirmed that high
binding energy as well as good molecular interactions with
important residues in the CLDN4 catalytic site. The ligand
NCI110039 docked with CLDN4 as a result −9.3 kcal/mol
of binding energy and H‐bonds with ARG31, ASN142,
ASP146, and ARG158 and hydrophobic interactions with
FIGURE 1 Structure of claudin‐4 with four transmembrane
domains, two extracellular loops with N‐ and C‐terminals. ECL1,
extracellular loop1; ECL2, extracellular loop2
The pictorial assessment disclosed that LEU176 is far
from the dynamic site region which does not exist in 5 Å
of the dynamic site. Furthermore, stereochemical ele-
ments of the CLDN4 were confirmed by verifying 3D web
server. An ideal 3D‐1D profile for each of the 20 amino
acids ought to be in the scope of 0 to 0.2. In the 3D server,
the values were lesser than zero, which are considered as
an erroneous for the model. The 3D plot of CLDN4
demonstrates that average of all the residues is 0.16,
which is generally near to 0.20. To assess the reliability of
the modeled structure of CLDN4, we calculated the root
mean square deviation (RMSD) by superimposing it on
the template using the Schrodinger. The CLDN4 model
shows 3.0 Å RMSD and an outstanding correspondence
with the CLDN15 (Mus musculus) experimental structure
44
TYR67, LEU71, LEU77, TRP138, and VAL152 (Figure 2).
Moreover, three pi‐pi interactions with ARG31 and ASP146.
The ligand NCI344682 docked with CLDN4 as a result
45
−
9.0 kcal/mol of binding energy and H‐bonds with
ARG31, ASP76, ASN142, ASP146, and ARG158 and
hydrophobic interactions with TYR67, LEU71, LEU77,
ILE143, and VAL152 (Figure 3). The NCI661251 ligand
with CLDN4 target docking was performed as a result
−9.6 kcal/mol of binding energy and H‐bonds with ARG31,
TYR67, ASP76, ASN142, and ASP146 and hydrophobic
interactions with TYR67, LEU71, LEU77, and ILE143
(Supporting Information Figure S2).
(Supporting Information Table S1 and Figure 4).
3.2
|
Virtual screening
3.4
|
Binding free energy analysis
The NCI database screened against the target binding site
of CLDN4 protein was utilized for virtual screening
workflow (Schrodinger, New York). The Glide HTVS was
done with default docking algorithm by selected con-
straints for every grid in the OPLS‐2005 force field. Using
Glide HTVS, around 9294 molecules were selected based
on GlideScore. After HTVS, the molecules having more
We have performed binding free energy studies to
validate molecular docking energy predictions. The
binding energy of NCI110039 is −24.341 kcal/mol,
NCI344682 is −33.947, and NCI661251 is −56.212 as
shown in Table 2. The MM/GBSA binding energies are
approximate free energies of binding, a more negative
energy value indicates stronger binding.
TABLE 1 Docking energies and interaction profile of claudin‐4 with NCI110039, NCI344682, and NCI661251
Type of interactions and residues involved
Glide docking
Hydrophobic interactions and pi‐pi
stackings
(
XP) energies,
H‐bonds
Sl no.
Ligand names
kcal/mol
1
NCI110039
−9.3
ARG31, ASN142, ASP146,
ARG158
TYR67, LEU71, LEU77, TRP138, VAL152
(Hphobic), ARG31, ASP146 (pi‐pi)
2
3
NCI344682
NCI661251
−9.0
−9.6
ARG31, ASP76, ASN142,
ASP146, ARG158
TYR67, LEU71, LEU77, ILE143, VAL152
ARG31, TYR67, ASP76,
ASN142, ASP146
TYR67, LEU71, LEU77, ILE143
Abbreviations: H‐bonds, hydrogen bonds; NCI, National Cancer Institute.

6
RAMBABU AND JAYANTHI
|
FIGURE 2 Docking snapshot of claudin‐4 with NCI110039. A, Showing hydrogen bond formed by NCI110039 with claudin‐4 residues.
B, 2D representation snapshot showing various interactions formed by NCI110039 at catalytical active site. NCI, National Cancer Institute
3
.5
|
Predicted ADME properties
were ended up being a strong criterion for evaluating drug‐
likeness. The above‐mentioned properties from QikProp
should be in the particular range such as octanol/water
partition coefficient (−2‐6.5), molecular weight (150‐500),
cell permeability (>500 is good, <25 is poor), brain/blood
partition (−3.0‐1.2), human oral absorption (≤25% is low, ≥
is high), solubility (−6.5‐0.5). The above‐mentioned
NCI110039, NCI344682, and NCI661251 structures are well
permissible range (Table 3).
ADME properties analyzed with QikProp with appropriate
and pharmaceutically noteworthy properties for the short-
listed three NCI molecules and the important properties
comprise of octanol/water partition coefficient, molecular
weight, Madin-Darby canine kidney (MDCK) cell perme-
ability, brain partition coefficient, human oral absorption,
aqueous solubility allowing to the Lipinski’s rule of five
FIGURE 3 Docking snapshot of claudin‐4 with NCI344682. A, Hydrogen bond formed by NCI344682 with claudin‐4 residues. B,
Two‐dimensional representation snapshot showing various interactions formed by NCI344682 at the catalytical active site of claudin‐4. NCI,
National Cancer Institute

RAMBABU AND JAYANTHI
7
|
FIGURE 4 Docking snapshot of claudin‐4 with NCI661251. A, Hydrogen bond formed by NCI661251 with claudin‐4 residues. B,
Two‐dimensional representation snapshot showing various interactions formed by NCI661251 at the catalytical active site of claudin‐4
TABLE 2 Binding free energy calculation for CLDN with
NCI110039, NCI344682, and NCI661251
fluctuations (RMSF) (Figure 6A), RMSD (Figure 5A), the
radius of gyration (ROG) (Figure 5B), total energy (Figure
5
D), and intramolecular H‐bonds (Figure 5C). The CLDN4
Sl no. Molecules ID ΔG, kcal/mol ΔGvdW, kcal/mol
RMSD backbone was observed to be fluctuating about 4.0 Å
on an average for the simulation. However, it is noticed
that the convergence after 3‐nanosecond of simulated time.
The RMSD averaged constant snaps recommended that the
structure of the CLDN4 is stable throughout the
1
2
3
110 039
344 682
661 251
−24.341
−33.947
−56.212
−14.954
−22.886
−29.712
50‐nanosecond simulation. Using the RMSF diagram
3
.6
|
Molecular dynamic simulations
(Figure 6A) of CLDN4, it is found that the main peaks of
fluctuation between 38 to 40 residues were more than
For better understanding, the results of CLDN4 were
subjected into protein in apo (no ligand) state first. Later,
CLDN4 with the lead compounds like NCI110039,
NCI344682, and NCI661251. The statistically critical out-
comes and trajectory analysis of the dynamics have been
represented (Supporting Information Table S2).
2.6 Å. At 152 amino acid, it is up to 5.2 Å and residues
between 200 to 209 up to 13.8 Å which is the most elevated
deviation amid MD simulations. Rest all residues were
observed to be entirely steady and under 3.0 Å. ROG is used
46
to find the compactness of the CLDN4. During the
CLDN4 apo state simulation measured ROG in the scope of
20.9 to 22.6 Å and with an average of 21.7 Å (Figure 5B).
3
.7
|
MD simulations of CLDN4
With the ROG information, CLDN4 protein is stable
47
membrane protein (apo form)
without major peaks present in the structure. CLDN4
holds an average 145 intramolecular H‐bonds inside the
scope 122 to 171 (Figure 5C). At last, studied the CLDN4’s
To inspect the stability and conformational changes of the
CLDN4 apo state we observed the root mean square
TABLE 3 ADME properties of NCI lead compounds
Octanol/water
Brain/blood
Molecular
weight
(mol_MW)
partition
coefficient (log solubility
Po/W)
Aqueous
Apparent MDCK
cell permeability
(QPPMDCK)
partition
coefficient
(QplogBB)
Percent
human oral
absorption
Rule
of five
Molecules
(QplogS)
NCI110039
NCI344682
NCI661257
264.191
321.33
−0.467
−3.243
−3.866
−1.728
−0.395
0.879
0.79
0.497
−2.468
−2.4446
−2.453
29.866
0
0
0
0
0
340.381
−0.079
Abbreviations: ADME, absorption, distribution, metabolism, and excretion; MDCK, Madin‐Darby canine kidney; NCI, National Cancer Institute.

8
RAMBABU AND JAYANTHI
|
FIGURE 5 RMSD, ROG, intrahydrogen bonds, total energy. A, MD simulation trajectory analysis RMSD of claudin‐4 protein in the
presence of no ligand and in complex with NCI110039, NCI344682, and NCI661251. B, MD simulation trajectory analysis ROG of claudin‐4
protein in the presence of no ligand and in complex with NCI110039, NCI344682, and NCI661251. C, MD simulation trajectory analysis
intrahydrogen bonds of claudin‐4 protein in the presence of no ligand and in complex with NCI110039, NCI344682, and NCI661251. D, MD
simulation trajectory analysis total energy of claudin‐4 protein in the presence of no ligand and in complex with NCI110039, NCI344682, and
NCI661251. MD, molecular dynamic; NCI, National Cancer Institute; RMSD, root mean square deviation; ROG, radius of gyration
energy in the most alleviated conformation and it was
noticed that the average −2889 kcal/mol (Figure 5D).
Because every single perception in RMSF, ROG, RMSD,
energy contributions, and intramolecular H‐bonds collec-
tively, a study can say that the CLDN4 structure is stable
throughout the simulation.
3.8
|
CLDN4‐NCI110039 molecular
dynamic simulations
The molecular dynamics of CLDN4‐NCI110039 complex was
executed to better comprehend the impact of NCI110039
with the CLDN4 binding site. The CLDN4‐NCI110039
complex had a binding affinity of −9.3 kcal/mol. The results
FIGURE 6 RMSF, Interhydrogen bonds. A, MD simulation trajectory analysis RMSF of claudin‐4 protein in the presence of no ligand
and in complex with NCI110039, NCI344682, and NCI661251. B, MD simulation trajectory analysis interhydrogen bonds of claudin‐4
protein in the presence of NCI110039, NCI344682, and NCI661251. CLDN4, claudin‐4; MD, molecular dynamic; NCI, National Cancer
Institute; RMSF, root mean square fluctuations

RAMBABU AND JAYANTHI
9
|
4
9
in Figure 5A shows that the RMSD is usual for the first
stable conformation.
The ROG of the CLDN4‐
26‐nanosecond. Although at the time of binding pocket
NCI344682 indicates that CLDN4 had a little elaborated
(Figure 5B) in the existence of NCI110039 through
continuing an average 22.2 Å within the range of 21.3
to 23.1 Å statistically. A reasonable expansion of CLDN4
can be detected around the middle and initiated after the
22‐nanosecond of the simulation. The CLDN4 structure
is as apparent as ROG associating with the RMSD, ROG
expansion could be the reason for additional stabilization
of the CLDN4‐NC344682 complex. We also considered
the intramolecular H‐bonds existing through the mole-
cular dynamics CLDN4‐NCI344682 complex and noticed
that it continues 137 average intramolecular H‐bonds in
the range 110 to 158, respectively. The energy required
for stable conformation of CLDN4‐NCI344682 and it has
been detected to continue −3335 kcal/mol average energy
(Figure 5D) which is very much minimized when
compared with CLDN4’s apo state average energy
−2889 kcal/mol, respectively. To end, approaching to
interH‐bonds between receptor and ligand throughout
the simulation average was 3.6 within the range of 0 to 12
maximum H‐bonds (Figure 6B).
opens (Figure 5A) ensued a raise up to 5.4 Å for a CLDN4‐
NCI1110039 complex. The CLDN4 RMSF considered with
the presence of NCI110039, which had been noticed the
most highly active amino acid motions in CLDN4 native
state and minimized (Figure 6A) in the existence of
NCI110039. These data are extremely braced to the solid
restraining and steady of NCI110039 on CLDN4 with a
48
comparison of CLDN4’s apo state residue fluctuations.
Fifty‐nanoseconds of dynamics were used in the current
study of protein‐ligand complex, which enables the most
49
stable conformation. The ROG of the CLDN4‐NCI110039
indicates CLDN4 had somewhat elaborated (Figure 5B) in
the existence of NCI110039 through continuing 21.5 Å as an
average within the range of 21.1 to 22.3 Å statistically. The
CLDN4 structure as apparent with ROG associating with the
RMSD, ROG expansion could be the reason for additional
stabilization of the CLDN4‐NCI110039. We also considered
the intramolecular H‐bonds exist through the simulation of
the CLDN4 in complex with NCI110039 and found out that
it is continuing with an average of 143 intramolecular
H‐bonds in the range of120 to 165, respectively. The
investigation of the energy required for stable conformation
of CLDN4‐NCI110039 and it has been detected to continue
3
.10
|
CLDN4‐NCI661251 molecular
dynamic simulations
−3146 kcal/mol average energy (Figure 5D) which is very
much minimized in comparing to CLDN4’s apo state average
energy of −2889 kcal/mol, respectively. Finally, approaching
the interH‐bonds between receptor and ligand throughout
the simulation of an average 2.6 within the range of 0 to 10
maximum H‐bonds (Figure 6B).
The CLDN4‐NCI661251 complex showed binding affinity
of −9.6 kcal/mol which was acquired by utilizing the
Glide module of Schrodinger. Subsequently, considered
RMSD for the CLDN4 complex with NCI661251. The
results in Figure 5A shows that the RMSD at the time of
binding pocket opens ensued in a decrease up to 3.2 Å for
the CLDN4‐NCI661251 complex. The CLDN4 RMSF is
considered with the presence of NCI661251, which had
been noticed the most highly active amino acid motions
in CLDN4 native state and minimized (Figure 6A) in the
existence of NCI661251. The data are highly supporting
to restraining and the steadying of NCI661251 on CLDN4
with a comparison of CLDN4’s apo state residue
3
.9
|
CLDN4‐NCI344682 molecular
dynamic simulations
The native state CLDN4 dynamics and CLDN4‐
NCI110039 complex were performed, the molecular
dynamics of CLDN4‐NCI344682 was executed to better
comprehend the impact of NCI344682 with the CLDN4
binding site. The NCI344682‐CLDN4 complex has a
binding affinity of −9.0 kcal/mol acquired by utilizing the
Glide module of Schrodinger. Subsequently, it is con-
sidered RMSD for the CLDN4 complex with NCI344682.
The results represented in Figure 5A shows that the
RMSD is usual for the first 12‐nanosecond. Although, at
the time of binding pocket opens (Figure 5A) ensued a
raise up to 5.8 Å for the CLDN4 complex. The CLDN4
RMSF considered with the presence of NCI344682 which
had been noticed the most highly active amino acid
motions in CLDN4 native state and minimized (Figure
4
9
fluctuations.
The ROG of the CLDN4‐NCI661251
indicates that CLDN4 had expanded (Figure 5B) in the
existence of NCI661251 through continuing 22.0 Å as an
average within the range of 21.5 to 22.8 Å statistically.
The CLDN4 structure as apparent with ROG associating
with the RMSD, ROG expansion could be the reason for
additional stabilization of the CLDN4‐NCI661251 com-
plex. The intramolecular H‐bonds exist throughout the
simulation and found out that it is continuing with an
average of 137 intramolecular H‐bonds in range 115 to
160 respectively, through the simulation representing
complex stability. The energy required for stable con-
formation of CLDN4‐NCI661251 and which had been
detected to continue with −3066 kcal/mol of average
6
A) in the existence of NCI344682. The ligand molecule
NCI344682 on CLDN4 with a comparison of CLDN4’s
apo state residue fluctuation, which enables the most

10
RAMBABU AND JAYANTHI
|
FIGURE 7 Total contacts, interaction diagram, torsional angle. A, Analysis of total contacts formed between claudin‐4 residues and
NCI110039 during MD simulations. B, Analysis of various interactions involved in stabilizing the claudin‐4‐NCI110039 complex. C, Analysis
of the torsional degree of freedom during MD simulation trajectory for the rotatable bonds present in the NCI110039. MD, molecular
dynamic
energy (Figure 5D) and is minimized when compared
with CLDN4’s apo state average energy −2889 kcal/mol
in the apo state. Further approaching interH‐bonds
between the CLDN4‐NCI661251 simulation average is
69% tenancy throughout MD trajectory. Aside from H‐
bonds, CLDN4‐NCI110039 complex shown two hydropho-
bic contacts with residues like TYR67, LEU77 (Table 4 and
Figure 7B).
6
.3 within the range starts from 0 to 19 maximum
To analyze and figure out the ligand torsions dynamics,
it enables interactions such as H‐bonds alongside different
H‐bonds (Figure 6B).
50
interactions amid CLDN4 and NCI110039 complex. The
calculated torsional degree of freedom for rotational bonds
in the NCI110039. NCI110039 molecule holds five
rotatable bonds amid ligand positions 1 to 5, 6 to 13, 7
to 19, 3 to 18, and 2 to 17. These rotational bonds help for
binding (Figure 7C).
Through a few computational examinations, it is found
that NCI110039 is the potential ligand and synthesized.
The synthesized ligand NCI110039 had been confirmed
with NMR analysis.
3
.11
|
CLDN4‐NCI110039 interaction
profile
The interaction profile incorporated inside the Schrodin-
ger’s Desmond module for canvassing the point by point
intermolecular interactions among CLDN4 and NCI110039
compound. A total of 5 to 10 (Figure 7A) H‐bonds were
existing among CLDN4 and NCI110039. Regular H‐bonds
were noticed with amino acids ARG31; ARG158 with over
TABLE 4 Interaction profile of NCI110039 with claudin‐4 complex post‐MD simulation comparison to pre‐MD simulations (ie,
molecular docking interactions)
S.no
Ligand name
Type of interactions
Pre‐MD
Post‐MD
H‐bond
ARG31, ASN142, ASP146, ARG158
TYR67, LEU71, LEU77, TRP138, VAL152
ARG31, ASP146
ARG31, ARG158
TYR67, LEU77, VAL152
N/A
1
NCI110039
Hydrophobic
Pi‐Pi
Abbreviations: MD, molecular dynamic; NCI, National Cancer Institute.

RAMBABU AND JAYANTHI
11
|
FIGURE 8 Structure of purpurogallin carboxylic acid (PCA), MCF7 cells viability, A549 cells viability. A, Structure of synthesized
molecule PCA. B, Purpurogallin carboxylic acid effect on MCF7 cell line. C, Purpurogallin carboxylic acid effect on A549 cell line
3
.12
|
1 H Nuclear magnetic resonance
concentration cytotoxicity had significantly increased (46.3%)
Figure 8C).
(
Ten grams yield 38%; orange crystals, melting point 320°C;
1
H NMR spectrum (400 MHz, DMSO‐d6) had shown the
peaks obtained at δ 6.706‐6.758 (q, 20 Hz, 1 H), δ 6.904 (s,
OH), δ 7.062‐7.085 (d, 8 Hz, 1 H), δ 7.331‐7.360 (d, 12 Hz,
4
|
CONCLUSION
1
H), δ 9.392 (s, 2 OH), δ 10.586 (s, OH), δ 15.321 (s, OH);
C NMR spectrum (400 MHz, DMSO‐d6): ppm 110.7,
15.3, 116.9, 124.0, 133.5, 134.8, 135.2, 152.0, 152.2, 155.2,
82.7. Calculated for C H O 264.0270; found 264.1320.
In the present study, homology modeling was carried out
to build an appropriate structure of CLDN4. The structure‐
based virtual screening was carried out to bring new and
potent inhibitors for CLDN4 protein. Further, docking
studies with compounds from the NCI database elucidated
the binding mechanism of CLDN4 and active residues like
ARG31, ASN142, ASP146, and ARG158 were identified.
Subsequently, MD simulations have been performed for
CLDN4‐NCI110039, CLDN4‐NCI344682, and CLDN4‐
NCI661251 complexes that provide insight into interacting
residues and protein stability. Moreover, molecular
dynamic simulations results showed that the trajectory
of the CLDN4 complex with NCI110039 is stable and
insignificant conformational change was observed in
CLDN4 because of NCI110039 binding. The stability
NCI110039 was determined by executing MM/GBSA and
MD simulation studies. In addition, preliminary in vitro
studies were also performed and the cytotoxicity assays
showed that NCI110039 has an inhibitory effect in the
breast (MCF7) and lung (A549) cancer cell lines. From the
overall analysis through molecular docking, dynamic
simulations and cytotoxic studies, the current study
1
3
1
1
12
8 7
3
.13
|
Biological activity of
purpurogallin carboxylic acid
The cytotoxic effect of purpurogallin carboxylic acid (Figure
8A) was determined by MTT assay through in vitro on
MCF7 and A549 cell lines. Cytotoxicity of purpurogallin
carboxylic acid was measured with different concentrations.
Results showed compare with the control, 1 µM had not
shown any cytotoxicity effect. With increasing concentration
of purpurogallin carboxylic acid, cytotoxicity also signifi-
cantly increased till 25 µM (31.9%). After increasing the
concentration (50 µM) it had not shown any cytotoxic effect
(38%). It indicates that the 25 µM is the optimal concentra-
tion for cytotoxicity of MCF7 cell line (Figure 8B). On A549
cell line with 1 and 2.5 µM cytotoxicity significantly increased
(81.2% and 61%). With 5, 10, and 25 µM it is shown a
moderate cytotoxicity (56.9%, 53.4%, and 52.5%). At 50 µM

12
RAMBABU AND JAYANTHI
|
tight junction barrier function in ovarian cancer cells. J Biol
Chem. 2005;280:26233‐26240.
1. D'souza T, Indig FE, Morin PJ. Phosphorylation of claudin‐4 by
PKCepsilon regulates tight junction barrier function in ovarian
cancer cells. Exp Cell Res. 2007;313:3364‐3375.
2. Kyuno D, Kojima T, Ito T, et al. Protein kinase Cα inhibitor
enhances the sensitivity of human pancreatic cancer HPAC
cells to Clostridium perfringens enterotoxin via claudin‐4. Cell
Tissue Res. 2011;346:369‐381.
reveals that NCI110039 could be a promising lead for
CLDN4 inhibitory action. Further CLDN4 knockout
studies are required in future to substantiate the mechan-
ism of action of NCI110039 to serve as a potent anticancer
agent.
1
1
ACKNOWLEDGMENTS
1
3. Li J, Chigurupati S, Agarwal R, et al. Possible angiogenic roles for
The authors are thankful for the financial support from
the Department of Science and Technology, Science and
Engineering Research Board (DST‐SERB), DST No:
SB/YS/LS‐299/2013. The authors would like to thank
the management of Vellore Institute of Technology,
Vellore for providing necessary support for carrying out
this study work.
claudin‐4 in ovarian cancer. Cancer BiolTher. 2009;8:1806‐1814.
4. Agarwal R, D'souza T, Morin PJ. Claudin‐3 and claudin‐4
expression in ovarian epithelial cells enhances invasion and is
associated with increased matrix metalloproteinase‐2 activity.
Cancer Res. 2005;65:7378‐7385.
5. Honda H, Pazin MJ, Ji H, Wernyj RP, Morin PJ. Crucial roles of
sp1 and epigenetic modifications in the regulation of the cldn4
promoter in ovarian cancer cells. J Biol Chem. 2006;281:21433‐
1
1
2
1444.
ORCID
16. Kwon MJ, Kim SS, Choi YL, et al. Derepression of cldn3 and cldn4
during ovarian tumorigenesis is associated with loss of repressive
histone modifications. Carcinogenesis. 2010;31:974‐983.
Sivaraman Jayanthi
658
http://orcid.org/0000-0001-6722-
1
17. Rangel LB, Agarwal R, D'souza T, et al. Tight junction proteins
claudin‐3 and claudin‐4 are frequently overexpressed in ovarian
cancer but not in ovarian cystadenomas. Clin Cancer Res. 2003;9:
Uncited reference
2567‐2575.
50
18. Boylan KLM, Misemer B, Derycke M, et al. Claudin 4 is
differentially expressed between ovarian cancer subtypes and
plays a role in spheroid formation. Int J Mol Sci. 2011;12:
.
1
334‐1358.
REFERENCES
1
9. Bellone S, Tassi R, Betti M, et al. Mammaglobin b (scgb2a1) is a
novel tumor rejection antigen and highly differentially ex-
pressed gene in all major histological types of ovarian cancer.
Br J Cancer. 2013;109:462‐471.
20. Patonai A, Erdélyi‐Belle B, Korompay A, et al. Claudins and
tricellulin in fibrolamellar hepatocellular carcinoma. Virchow-
sArchiv. 2011;458:679‐688.
1
. Furuse M, Sasaki H, Fujimoto K, Tsukita S. A single gene
product, claudin‐1 or ‐2, reconstitutes tight junction strands and
recruits occludin in fibroblasts. J Cell Biol. 1998;143:391‐401.
. Tsukita S, Furuse M. Pores in the wall: claudins constitute
tight junction strands containing aqueous pores. J Cell Biol.
2
2
000;149:13‐16.
3
4
5
. Morin PJ. Claudin proteins in human cancer: promising new
targets for diagnosis and therapy. Cancer Res. 2005;65:9603‐9606.
. Singh AB, Sharma A, Dhawan P. Claudin family of proteins
and cancer: an overview. J Oncol. 2010;2010:541957‐541968.
. Ohtani S, Terashima M, Satoh J, et al. Expression of tight‐
junction‐associated proteins in human gastric cancer: down-
regulation of claudin‐4 correlates with tumor aggressiveness
and survival. Gastric Cancer. 2009;12:43‐51.
. Sung CO, Han SY, Kim SH. Low expression of claudin‐4 is
associated with poor prognosis in esophageal squamous cell
carcinoma. Ann Surg Oncol. 2011;18:273‐281.
. Ueda J, Semba S, Chiba H, et al. Heterogeneous expression of
claudin‐4 in human colorectal cancer: decreased claudin‐4
expression at the invasive front correlates cancer invasion and
metastasis. Pathobiology. 2007;74:32‐41.
21. Aravindakshan J, Chen X, Sairam MR. Differential expression
of claudin family proteins in mouse ovarian serous papillary
epithelial adenoma in aging FSH receptor‐deficient mutants.
Neoplasia. 2006;8:984‐994.
22. NCI, https://cactus.nci.nih.gov/download/nci/index.html. Ac-
cessed May 2012.
23. Suzuki H, Nishizawa T, Tani K, et al. Crystal structure of a
claudin provides insight into the architecture of tight junctions.
Science. 2014;344:304‐307.
24. Jorgensen WL, Maxwell DS, Tirado‐Rives J. Development and
testing of the OPLS all‐tom force field on conformational
energetics and properties of organic liquids. J Am Chem Soc.
1996;118:11225‐11236.
25. Kaminski GA, Friesner RA, Tirado‐Rives J, Jorgensen WL.
Evaluation and reparameterization of the OPLS‐AA force
field for proteins via comparison with accurate quantum
chemical calculations on peptides. J Phys Chem B. 2001;105:
6474‐6487.
6
7
8
. Shield K, Ackland ML, Ahmed N, Rice GE. Multicellular
spheroids in ovarian cancer metastases: biology and pathology.
Gynecol Oncol. 2009;113:143‐148.
9
. Boyle P, Levin B. World Cancer Report 2008. France: Interna-
26. Jorgensen WL. The many roles of computation in drug
tional Agency for Research on Cancer; 2008.
0. D'Souza T, Agarwal R, Morin PJ. Phosphorylation of claudin‐3
discovery. Science. 2004;303:1813‐1818.
27. Song CM, Lim SJ, Tong JC. Recent advances in computer‐aided
drug design. Brief Bioinform. 2009;10:579‐591.
1
at threonine 192 by cAMP‐dependent protein kinase regulates

RAMBABU AND JAYANTHI
13
|
2
2
3
8. Leach AR, Shoichet BK, Peishoff CE. Prediction of protein‐
ligand interactions. Docking and scoring: successes and gaps. J
Med Chem. 2006;49:5851‐5855.
9. Tirado‐Rives J, Jorgensen WL. Contribution of conformer
focusing to the uncertainty in predicting free energies for
protein‐ligand binding. J Med Chem. 2006;49:5880‐5884.
0. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experi-
mental and computational approaches to estimate solubility
and permeability in drug discovery and development settings.
Adv Drug Deliv Rev. 2001;46:3‐26.
free’apoptotic condition in part via the Dor/pi3k/akt pathway.
Life Sci. 2017;191:195‐204.
43. Bhavaniprasad V, Febin prabhu dass J, Jayanthi S. Activation
mechanism of claudin‐4 by ephrin type‐A receptor 2: a molecular
dynamics approach. Mol Biosyst. 2013;9:2627‐2634.
44. Venkatesan A, Rambabu M, Jayanthi S, Febin Prabhu Dass J.
Pharmacophore feature prediction and molecular docking
approach to identify novel anti‐HCV protease inhibitors. J Cell
Biochem. 2018;119:960‐966.
45. Pandey B, Sharma P. Structural insights into impact of
Y134F mutation and discovery of novel fungicidal com-
pounds against CYP51 in Puccinia triticina. J Cell Biochem.
2018;119:2588‐2603.
31. Mariadasse R, Biswal J, Jayaprakash P, et al. Mechanical insights of
oxythiamine compound as potent inhibitor for human transketo-
lase‐like protein 1 (TKTL1 protein). J Recept Signal Transduct Res.
2
016;36:233‐242.
46. Lobanov MY, Bogatyreva NS, Galzitskaya OV. Radius of gyration
as an indicator of protein structure compactness. J Mol Biol.
2008;42:623‐628.
47. Purohit R. Role of ELA region in auto‐activation of mutant KIT
receptor: a molecular dynamics simulation insight. J Biomol
Struct Dyn. 2014;32:1033‐1046.
3
2. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental
and computational approaches to estimate solubility and perme-
ability in drug discovery and development settings. Adv Drug Deliv
Rev. 1997;23:3‐25.
3
3. Jorgensen WL, Duffy EM. Prediction of drug solubility from monte
carlo simulations. Bioorg Med Chem Lett. 2000;10:1155‐1158.
4. Jorgensen WL, Duffy EM. Prediction of drug solubility from
48. Kumar A, Purohit R. Computational investigation of patho-
genic nsSNPs in CEP63 protein. Gene. 2012;503:75‐82.
49. DuBay KH, Geissler PL. Calculation of proteins’ total side‐
chain torsional entropy and its influence on protein–ligand
interactions. J Mol Biol. 2009;391:484‐497.
50. Reddy SVG, Reddy KT, Kumari VV, Basha SH. Molecular
docking and dynamic simulation studies evidenced plau-
sible immunotherapeutic anticancer property by Withaferin
A targeting indoleamine 2, 3‐dioxygenase. J Biomol Struct
Dyn. 2015;33:2695‐2709.
3
structure. Adv Drug Deliv Rev. 2002;54:355‐366.
3
5. Bowers KJ, Chow E, Xu H et al. Scalable algorithms for molecular
dynamics simulations on commodity clusters. In SC 2006 conference:
proceedings of the ACM/IEEE; Heraklion, Greece, 2006.
6. Schrödinger Release 2014‐2: Desmond Molecular Dynamics
System, version 3.8, D. E. Shaw Research, New York, NY, 2014.
Maestro‐Desmond Interoperability Tools, version 3.8, Schrö-
dinger, New York, NY, 2014.
3
3
7. Shivakumar D, Williams J, Wu Y, Damm W, Shelley J,
Sherman W. Prediction of absolute solvation free energies
using molecular dynamics free energy perturbation and the
OPLS force field. J Chem Theory Comput. 2010;6:1509‐1519.
8. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein
ML. Comparison of simple potential functions for simulating
liquid water. J Chem Phys. 1993;79:926‐935.
9. Shinoda W, Mikami M. Rigid‐body dynamics in the isothermal‐
isobaric ensemble: a test on the accuracy and computational
efficiency. J Comput Chem. 2003;24:920‐930.
SUPPORTING INFORMATION
3
3
4
Additional supporting information may be found online
in the Supporting Information section at the end of the
article.
0. Jhoo JW. Anti‐inflammatory effects of purpurogallin carboxylic
acid, an oxidation product of gallic acid in fermented tea.
Korean J Food Sci Technol. 2008;40:707‐711.
How to cite this article: Rambabu M, Jayanthi S.
Virtual screening of National Cancer Institute
database for claudin‐4 inhibitors: Synthesis,
biological evaluation, and molecular dynamics
studies. J Cell Biochem. 2018;1‐13.
4
1. Crow WD, Haworth RD. 293. Purpurogallin. Part VII. Constitution
of purpurogallincarboxylic acid. J Chem Soc (Resumed). 1951;131:
325‐1327.
1
42. Reddy LVK, Sen D. DADLE enhances viability and anti‐
https://doi.org/10.1002/jcb.28147
inflammatory effect of human Mscs subjected to ‘serum