Design of Novel Selective Estrogen Receptor Inhibitors using Molecular Docking and Protein- Ligand Interaction Fingerprint Studies

Abstract

Aims: The genomic and non-genomic actions of human estrogen receptor α (hERα) play a crucial role in breast epithelial cell proliferation and survival, as well as mammary tumorigenesis. hERα has been proved as a potential target for breast cancer therapy over the last 3 decades. Hence designing novel inhibitors targeting hERα can be a valuable approach in breast cancer therapy. Study Design: In the present study, the goal is to identify novel hERα inhibitors through molecular docking, AI based virtual screening and interaction fingerprint analysis. Place and Duration of Study: Department of Bioinformatics, Sri Venkateswara Institute of Medical Sciences, Tirupati, Andhra Pradesh, India in between July 2021-sep 2021. Methodology: Molecular docking studies were performed using the human estrogen receptor alpha ligand-binding domain in complex with 4-hydroxytamoxifen (PDB: 3ERT) against existing compounds from literature. The best docked existing compound and co-crystal ligand were subjected to shape screening against 28 million compounds and resulted compounds constituted Original Research Article Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474 471 the hERα inhibitor dataset which was subjected to rigid receptor docking. Further, interaction fingerprint analysis was applied as complimentary method to docking for comparing the similarity of predicted binding poses of proposed leads with that of reference binding pose. Results: Co-crystal ligand (4-OHT) and E99 exhibited better binding affinity among existing ligand set. Rigid receptor docking studies resulted in four lead compounds possessing better docking scores than 4-OHT and E99. Moreover, leads showed favorable absorption, distribution, metabolism, excretion and toxicity properties within the range of 95% FDA approved drugs. Proposed leads showed interactions with binding site residues of hERα similar to that of 4-OHT with better binding affinity. The ability of obtained leads to retrieve actives was validated using receiver operative characteristic (ROC) curve. Conclusion: From above results, it has been observed that the proposed leads have potential to act as novel hERα inhibitors.

Full text

_____________________________________________________________________________________________________
*Corresponding author: E-mail: rajitha.galla@gmail.com;
Journal of Pharmaceutical Research International
33(46A): 470-483, 2021; Article no.JPRI.75474
ISSN: 2456-9119
(Past name: British Journal of Pharmaceutical Research, Past ISSN: 2231-2919,
NLM ID: 101631759)
Design of Novel Selective Estrogen Receptor
Inhibitors using Molecular Docking and Protein-
Ligand Interaction Fingerprint Studies
Galla Rajitha1*, Murthi Vidya Rani1, Umakanth Naik Vankadoth2
and Amineni Umamaheswari2
1Institute of Pharmaceutical Technology, Sri Padmavati Mahila Visvavidyalayam, Tirupati-517502,
Chittoor Dist, Andhra Pradesh, India.
2Department of Bioinformatics, Sri Venkateswara Institute of Medical Sciences, Tirupati-517502,
Chittoor Dist, Andhra Pradesh, India.
Authors’ contributions
This work was carried out in collaboration among all authors. All authors read and approved the final
manuscript.
Article Information
DOI: 10.9734/JPRI/2021/v33i46A32890
Editor(s):
(1) Dr. Rafik Karaman, Al-Quds University, Palestine.
Reviewers:
(1) Mohamad Zulkeflee bin Sabri, Universiti Kuala Lumpur Malaysian Institute of Chemical and Bioengineering Technology,
Malaysia.
(2) Anjali M. Wanegaonkar, University of Mumbai, India.
Complete Peer review History: https://www.sdiarticle4.com/review-history/75474
Received 03 August 2021
Accepted 11 October 2021
Published 16 October 2021
ABSTRACT
Aims: The genomic and non-genomic actions of human estrogen receptor α (hERα) play a crucial
role in breast epithelial cell proliferation and survival, as well as mammary tumorigenesis. hERα
has been proved as a potential target for breast cancer therapy over the last 3 decades. Hence
designing novel inhibitors targeting hERα can be a valuable approach in breast cancer therapy.
Study Design: In the present study, the goal is to identify novel hERα inhibitors through molecular
docking, AI based virtual screening and interaction fingerprint analysis.
Place and Duration of Study: Department of Bioinformatics, Sri Venkateswara Institute of Medical
Sciences, Tirupati, Andhra Pradesh, India in between July 2021-sep 2021.
Methodology: Molecular docking studies were performed using the human estrogen receptor
alpha ligand-binding domain in complex with 4-hydroxytamoxifen (PDB: 3ERT) against existing
compounds from literature. The best docked existing compound and co-crystal ligand were
subjected to shape screening against 28 million compounds and resulted compounds constituted
Original Research Article

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
471
the hERα inhibitor dataset which was subjected to rigid receptor docking. Further, interaction
fingerprint analysis was applied as complimentary method to docking for comparing the similarity of
predicted binding poses of proposed leads with that of reference binding pose.
Results: Co-crystal ligand (4-OHT) and E99 exhibited better binding affinity among existing ligand
set. Rigid receptor docking studies resulted in four lead compounds possessing better docking
scores than 4-OHT and E99. Moreover, leads showed favorable absorption, distribution,
metabolism, excretion and toxicity properties within the range of 95% FDA approved drugs.
Proposed leads showed interactions with binding site residues of hERα similar to that of 4-OHT
with better binding affinity. The ability of obtained leads to retrieve actives was validated using
receiver operative characteristic (ROC) curve.
Conclusion: From above results, it has been observed that the proposed leads have potential to
act as novel hERα inhibitors.
Keywords: Human estrogen receptor-α; artificial intelligence; molecular docking; interaction
fingerprints; virtual screening; ligand dataset.
1. INTRODUCTION
Breast cancer refers to the erratic growth and
proliferation of cells which undergo changes in
their molecular characteristics that originate in
the breast tissue [1]. Breast cancer accounts for
14.8% of cancers in Indian women, it is reported
that an Indian woman is being diagnosed with
breast cancer for every four minutes, and is one
of the leading causes of cancer death in women
[2-4]. Frequently occurring breast cancers are
infiltrating ductal carcinoma, infiltrating lobular
carcinoma; lobular carcinoma in situ; ductal
carcinoma in situ [5].
Breast cancers were categorized into 4 types
using genetic information about breast cancer
cells. These breast cancer groups include: a)
Group-1 (luminal A) tumors that are estrogen
receptor (ER) positive and progesterone receptor
(PR) positive, but human epidermal growth factor
receptor 2 (HER2) negative. b) Group-2 (luminal
B) tumors are ER positive, PR negative and
HER2 positive. C) Group-3 (HER2 positive)
tumors are ER negative and PR negative, but
HER2 positive. D) Group-4 (basal-like) is also
known as triple-negative breast cancer, that are
ER negative, PR negative and HER2 negative
[6]. As stated by American Cancer Society, about
2 out of 3 cases of breast cancer are hormone
receptor-positive [7]. Among them ERα-positive
breast cancer is the most common type (70%).
ER-α, a major subtype of ER, plays crucial role in
breast cancer cell proliferation, invasion and
apoptosis [8]. Most ERα ligands target the ligand
binding domain. Two distinct activation functions
(AFs), AF-1 of N-terminus, AF-2 in the ligand
binding domain promotes transcriptional
activation by ERα. AF-1 activity is regulated by
growth factors involved in the Mitogen-activated
protein kinase pathway, while AF-2 function is
responsive to the ligand binding i.e. agonists
stimulate AF-2 activity while antagonists does not
stimulate AF-2 activity [9].
Selective estrogen receptor modulators (SERMs)
are used for preventing the effect of estrogens in
breast tissues. These compounds bind to the ER
similar to natural estrogens so that estrogen
cannot attach to a breast cell, thus preventing
estrogen's signals to cell to grow and multiply
[10]. Currently available SERMs include
raloxifene, lasofoxifene and bazedoxifene [11].
Known SERM tamoxifen binds with ER and
prevents growth of breast tissue [12].
Unfortunately, it increases the risk of blood clots
in legs and lungs and also increases the risk of
developing endometrial cancer [13]. Aromatase
inhibitors like exemestane, anastrozole
and letrozole inhibit aromatase thus minimizes
estrogen production. But, these drugs
associated with risk of fractures and
osteoporosis. Chemotherapy may also cause
various side effects and suppress the
production of blood cells causing
anemia/bleeding or weakens immune system
[14]. Hence, there is a need for development of
novel safe and effective anti-estrogenic agents.
Machine learning and artificial intelligence-based
methods for ligand discovery have received a
significant attention in both academic research
and industry due to high cost of clinical trials and
low success rate (6.2%), to minimize the cost
and also to identify potential therapeutic agents
[15]. In modern drug discovery, computational
techniques such as molecular docking and virtual
screening approaches minimizes the initial level
of research to discover potential drug candidates
for a disease [16]. Deep learning algorithms can

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
472
understand the patterns within existing dataset
and play vital role with virtual screening. The
drug discovery process applies algorithms to
understand the pattern of different chemical
compounds with diseases [17]. Interaction
fingerprints are binary representations of protein-
ligand complexes and these are valuable tools in
virtual screening to identify novel hits that can be
eventually optimized to drug candidates [18].
As there are no ideal antiestrogenic agent
available for breast cancer therapy, the present
study focused on designing novel estrogen
receptor inhibitors using molecular docking,
artificial intelligence based virtual screening,
interaction fingerprint and receiver operative
characteristic (ROC) metrics.
2. MATERIALS AND METHODS
2.1 Datasets
2.1.1 Ligand dataset collection and
preparation
Selective estrogen receptor inhibitors were
retrieved from literature. Structure of co-crystal
ligand 4-hydroxytamoxifen (4-OHT) was retrieved
from PDB database. Ligand preparation was
done using LigPrep application with Epik module
of Schrödinger, including enhancement of ligand
stereochemical nature, protonation states,
developing tautomers and finally energy
minimization of the ligand using the force field
OPLS_3 at pH7.0 ± 2.0 [19,20].
2.1.2 Protein preparation
Among the available crystal structures, the best
resolute X-ray structure of human estrogen
receptor alpha (hERα) ligand-binding domain co-
crystallized with 4-hydroxytamoxifen (PDB:
3ERT) was considered in this study to propose
antagonists through virtual screening, molecular
docking and protein-ligand interaction fingerprint
analysis. Estrogen receptor crystal structure was
imported to Maestro and then prepared using
protein preparation wizard of the Schrödinger by
addition of hydrogen atoms, bond order and
formal charge corrections, removal of atomic
clashes, tautomeric alterations and ionization
states of protein. The hydrogen bonding network
was optimized by reorienting the hydroxyl and
thiol groups in the protein and performed other
operations that are not part of the X-ray crystal
structure refinement process. Optimization of
protein was done at neutral pH and then the
energy minimization was done by applying
optimized potentials for liquid simulations (OPLS-
3) force field for all atoms. A receptor grid was
generated around inhibitor binding site residues
of the estrogen receptor using Glide v7.1. Using
the grid region, the undesirable water molecules
were removed from the inhibitor binding site of
target protein using Protein preparation wizard
[21].
2.2 Glide XP Docking
Docking is a procedure to predict the preferable
binding orientation between the two molecules to
form a stable complex. There are many steps in
which artificial intelligence methods come into
play in the docking, such as feature selection &
extraction, classification & regression for the
design of scoring functions and recognition of
binding sites. Few examples include applications
of probabilistic Naïve Bayes methods to improve
docking scoring functions, applications of neural
networks to virtual screening in combination with
docking etc. [22]. GLIDE (grid-based ligand
docking with energetics) XP (extra precision)
docking procedure was adopted for analyzing the
binding affinity between the protein and ligand.
The prepared and optimized ligands were flexibly
docked into the grid box generated around
inhibitor binding site residues of the protein using
Monte Carlo-based simulated algorithm
minimization method [23]. Glide Score (Gscore)
was used for representing binding affinity,
binding orientation and ranking. Docking was
implemented to retain the best molecules with
better binding affinity and good binding
orientation without steric clashes. 10 poses were
generated during XP docking for each ligand and
the best pose was retained after post-docking
minimization.
2.3 Exploring in-House Library
Based on the XP Gscore, the best docked
existing compound and co-crystal ligand were
screened against inhouse library molecules
containing more than 28 million compounds of
eMolecules, ChemBank, ChemPDB, KEGG
ligand, Unannotated NCI, Anti-HIV NCI, Drug
likeness NCI, AkosGmbh, Asinex, and TimTec
databases [24]. The best docked existing
compound and co-crystal ligand were imported
as hERα inhibitor dataset, to Maestro v11.1 for
virtual screening workflow to dock with hERα
using Glide with defined pH range 7.0±2.0 by
applying Qik Prop v5.1, Lipinski’s filter and
reactive filter. The best docked existing

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
473
compound, co-crystal ligand and resulted
compounds from shape screening against 28
million compounds constituted the hERα inhibitor
dataset.
2.4 Virtual Screening
Human ERα inhibitor dataset was docked with
inhibitor binding site residues of estrogen
receptor. Schrödinger virtual screening workflow
uses three flexible docking methods, namely
Glide high throughput virtual screening (HTVS),
standard precision (SP) and extra precision (XP)
docking [25]. The top ligand molecules obtained
through XP docking were compared with the co-
crystal ligand to propose novel hERα inhibitors
which constitute proposed leads. The obtained
leads were analyzed for pharmacological
descriptors and ADME/T properties.
2.5 Generation of Interaction Fingerprints
Docking interactions of proposed leads were
further evaluated by interaction fingerprint
analysis to observe whether leads exhibited
similar interactions when compared with that of
co-crystal ligand. Interaction fingerprint was
generated for the proposed leads and co-crystal
ligand docked complexes that translates 3D
structural binding information from a protein-
ligand complex into a one-dimensional binary
string [26]. Each fingerprint represents the
“structural interaction profile” of the complex that
can be used to organize, analyze, and visualize
the rich amount of information encoded in ligand
receptor complexes. Value 1 represents the
given interaction is established and 0 denotes
absence of specific interaction [18].
2.6 Evaluation of Virtual Screening
Based on docking scores and interaction
fingerprint analysis, the proposed leads were
further validated by enrichment factor (EF)
metrics. An internal library containing 1005
compounds was created with 1000 decoys from
Schrödinger along with proposed leads, and the
best co-crystal ligands among existing ligands of
hERα as actives. The discriminative ability of the
generated leads was evaluated by distinguishing
the active compounds from the internal library
consisting both actives and decoys. The leads
and co-crystal ligand were screened against the
internal library. The ligand dataset was docked to
the hERα and the results were analyzed using
EF at N% of sample size and BEDROC (α 20)
metrics for validating whether the best docked
compounds are reliable in picking the leads as
actives [25].
3. RESULTS AND DISCUSSION
3.1 Datasets
3.1.1 Ligand sets
The structures of 129 existing estrogen receptor
inhibitors were retrieved from literature,
and co-crystal ligand 4-OHT structure was
retrieved from PDB database for ligand set
preparation. All the ligands were prepared using
Lig Prep application with Epik module of
Schrödinger.
3.1.2 Protein preparation and binding site
analysis
The ideal protein 3ERT was selected to define
the antagonist behavior of leads because 3ERT
has ligand binding domain to show antagonist
behavior of 4-OHT. 4-OHT is the active
metabolite of tamoxifen which blocks the AF-2
activity by disrupting the topography of the AF-2
surface in hERα thus preventing transcriptional
activation. The energies of prepared crystal
structure of hERα were minimized by protein
preparation wizard. Inhibitor binding site residues
were defined around grid generated within the
hERα protein for further study as these residues
contribute to the structural and functional
properties of hERα. The hERα - 4-OHT complex
inhibitor binding site comprises residues such as
Met 343, Leu 346, Thr 347, Ala 350, Glu 353, Trp
383, Leu 384, Leu 387, Arg 394 and Leu 525
within the 4 Å region surrounding 4-OHT. The
binding site residues of hERα were defined 4 Å
around co-crystallized inhibitor using PD Bsum
[27].
3.2 Identification of the Best Existing
Ligand
About 129 existing ligands from literature and
one co-crystal ligand 4-hydroxytamoxifen (4-
OHT) were docked with hERα binding site
residues. Among all the docked complexes, 4-
OHT possesses least XP Gscore of -11.897
Kcal/mol with good bonded and non-bonded
interactions with inhibitor binding site residues
followed by existing ligands E99 with XP Gscores
-9.608 Kcal/mol (Table 1).

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
474
Table 1. Docking scores of hERα with proposed leads, best co-crystal ligand 4-OHT and best
docked existing ligands
S. No
Compound
XPG Score (Kcal/Mol)
1.
Lead 1
-12.668
2.
Lead 2
-12.19
3.
Lead 3
-11.927
4.
Lead 4
-11.899
5.
4-OHT
-11.897
6.
E99
-9.608
3.3 Virtual Screening
The best docked existing ligand E99 and co-
crystal ligand 4-OHT were chosen for screening
against in-house library database by applying
QikProp v5.1, Lipinski’s filter and reactive filter.
Lipinski’s filter was used to remove the
compounds which are not obeying Lipinski’s rule
of five and reactive filter was used to remove
ligands with reactive functional groups. Shape
screening resulted in 800 structurally similar
compounds. 800 structural analogues obtained
from the in-house library, best docked existing
ligand and co-crystal ligand were imported to
form an hERα inhibitor dataset for docking. Rigid
receptor docking method was streamlined
through HTVS, SP, and XP docking methods to
find potential leads. In virtual screening a large
number of molecules are ranked according to
their likelihood to be bioactive compounds, with
the aim to enrich the top fraction of the resulting
list. Out of 802 ligands docked in HTVS method,
top ranked ligands were re-docked using SP
method. 80 ligands obtained through SP method
were re-docked using XP docking, finally 10
docked complexes were obtained for each ligand
with docking scores. The docking strategy results
revealed that four leads showed the highest
binding affinity towards hERα than the best
docked existing ligand E99 and co-crystal ligand
4-OHT. Among hERα inhibitor dataset, lead 1
showed the highest binding affinity towards
hERα with XP Gscore of -12.668 kcal/ mol when
compared to that of co-crystal ligand -11.897
Kcal/Mol and the best docked compound E99
among 129 existing compounds -9.608 Kcal/mol
(Table 1). The molecules obtained from virtual
screening were subjected to predict the ADME
properties using QikProp module of Schrödinger
suite. In addition, SASA (solvent accessible
surface area gives information about transferring
free energy of bio molecule when it moves from
polar medium to nonpolar medium, SASA
indicates the accessible surface of the protein
solvent or a part of a protein that exposes to the
solvent), FOSA (Hydrophobic components of the
SASA), FISA (Hydrophilic components of the
SASA), and PISA (π carbon and attached
hydrogen components of the SASA) values of
docked ligands were carried using Schrödinger
suite [28]. Results indicated that leads showed
favorable ADME/T properties within the range of
95% FDA approved drugs (Tables 2 & 3).
3.4 Interactions of 4-OHT, Proposed
Leads and E99 with hERα
4-OHT formed two hydrogen bonds with side
chain residues of Glu 353 and Arg 394 (Fig. 1).
RRD strategy revealed that lead 1 conformation
is having better binding affinity than the co-
crystal ligand and obtained leads. The lead1
showed good binding affinity due to various
interactions with inhibitor binding site such as
hydrogen bonding, polar, hydrophobic,
electrostatic and stearic interactions with key
interacting residues with XPGscore of -12.668
Kcal/Mol [25]. Lead1
[Oc1ccc(cc1)[C@H](c1ccc(O)cc1)C(c1ccc(OCC[
NH+](CCO)C(CO)(CO)CO)cc1)c1ccc(O)cc1]
formed four hydrogen bonds, in which one
hydrogen bond observed with side chain residue
of Glu 353 and three hydrogen bonds were
observed with backbone residues of Ala 350, Met
528, and Leu 387. Lead 1 also involved in two pi-
pi stacking interactions with Trp 383 (Fig. 2).
Literature study revealed that ligand-protein
complexes stabilized by multiple aromatic
interactions involving tryptophan residue and it
was found that pi-pi stacking is essential for the
favorable electron correlation, whereas cation-π
contacts produce further electrostatic
contributions. [30]. Lead 2
[Oc1ccc(cc1)[C@H](c1ccc(O)cc1)C(c1ccc(OCC[
N@@H+](CCO)C(CO)(CO)CO)cc1)c1ccc(O)cc1]
formed six hydrogen bonds with binding site
residues in which two hydrogen bonds were
observed with side chain residues of Glu 353;
Arg 394 and four hydrogen bonds were observed
with backbone residue of Ala 350, Leu 387, Met

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
475
528 and Val 534. Lead 2 also involved in one pi-
pi stacking interaction with Trp 383 (Fig. 3). Lead
3
[Oc1ccc(cc1)[C@H](c1ccc(O)cc1)C(c1ccc(OCC[
N@H+](CCO)C(CO)(CO)CO)cc1)c1ccc(O)cc1]
formed one hydrogen bond with side chain
residue of Glu 353; three hydrogen bond
interactions with backbone residues of Ala 350,
Met 528, Val 534 and one pi-pi stacking
interaction with Trp 383 (Fig. 4). Lead 4
[CC1(C)C[C@@H](C)C[C@@](C1)(c1ccc(O[C@
@H](C)C[NH+](C)C)cc1)c1ccc(O)cc1]
formed two hydrogen bonds with side chain
residues of Glu 353 and Arg 394 similar to that of
co-crystal ligand 4-OHT (Fig. 5). Among existing
ligand dataset, E99 [[O-
]C(=O)/C=C\c1ccc(cc1)Nc1nc(nc(n1)c1ccc(O)cc
1)c1ccccc1] showed H-bond interaction with side
chain residue of Arg 394 and H-bond interaction
with backbone residue of Leu 387 (Fig. 6).
Fig. 1. Docking interactions of 4-OHT with hERα
Fig. 2. Docking interactions of lead 1 with hERα

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
476
Table 2. Pharmacological descriptors of proposed leads, best co-crystal ligand 4-OHT and best docked existing ligand
Compound
M.W
SASA
FOSA
FISA
WPSA
PISA
Vol
HBD
HBA
IP
Glob
Lead 1
589.684
920.44
201.54
297.45
0
421.45
1778.098
7
10.8
8.58
0.771152
Lead 2
589.684
854.641
188.74
278.374
0
387.527
1702.582
7
10.8
8.575
0.806839
Lead 3
589.684
910.973
196.127
308.347
0
406.5
1764.398
7
10.8
8.727
0.775158
Lead 4
395.584
735.876
488.896
59.833
0
187.147
1373.322
1
3.5
8.849
0.811973
4-OHT
387.521
746.383
308.481
61.301
0
376.602
1342.05
1
3.5
8.655
0.788344
E99
410.431
730.451
46.111
182.852
0
501.487
1282.963
3
6.25
8.909
0.781718
* MW = Molecular weight (130.0/725.0); SASA = Total solvent accessible surface area (300.0/1000.0); FOSA = Hydrophobic solvent accessible surface area (0.0/750.0); FISA
(Hydrophilic components of the SASA), WPSA= Weakly polar solvent accessible surface area (0.0/175.0); PISA = Carbon Pi solvent accessible surface area (0.0/450.0); Vol =
Molecular volume (A^3) (500.0/2000); HBD = Hydrogen bond Donor (0.0/6.0); HBA = Hydrogen bond acceptor (2.0/20.0); IP (eV) = Ionization potential (7.9/10.5); Glob =
Globularity (0.75/0.95)

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
477
Table 3. Predicted ADME/T properties of proposed leads, best co-crystal ligand 4-OHT and
best docked existing ligand
Compounds
Log P o/w
Log BB
Log KP
Log KhSa
Rule of 5
Rule of 3
Lead 1
2.821
-3.699
-5.467
0.009
2
2
Lead 2
2.527
-3.215
-5.235
-0.106
2
2
Lead 3
2.653
-3.785
-5.721
-0.016
2
2
Lead 4
5.739
0.056
-3.546
1.478
1
1
4-OHT
5.744
-0.298
-2.425
1.172
1
1
E99
4.217
-1.914
-2.355
0.258
0
1
*Log P o/w = log P for octanol/water (1.8/6.5); Log BB = log BB for brain/blood (-3.0/1.2); Log KP = log KP for
skin permeability (KP in cm/h); Log KhSa = serum protein binding (-2.5/1.5); Rule of 5: Lipinski’s Rule of 5
violations (maximum is 4); Rule of 3: Jorgensen Rule of 3 violations (maximum is 3) [29]
Fig. 3. Docking interactions of lead 2 with hERα
3.5 Interaction Fingerprint Analysis
A 9 bit interaction fingerprint was generated to
describe 3D protein-ligand interaction of
proposed leads and 4-OHT in the inhibitor
binding site of hERα. Each bit of the fingerprint
represents a pharmacophore feature and has 0
or 1, which means presence or absence of the
features in the conformation (Table 4) and
compared with 4-OHT. The interaction of
estradiol directly with key amino acid residues
such as Glu353 and Arg 394 in hERα is a major
determinant of ER’s specific recognition of
estrogens and thus estrogen promotes the
growth of the cancer. Tamoxifen (SERM), mimics
the shape and chemical composition of estradiol
and binds in the same site on the hERα as the
normal hormone but does not activate it, thus
prevents estrogen-induced growth. In 4-OHT-
ligand binding domain complex 4-OHT acts as an
antagonist and blocks the AF-2 activity by
disrupting the topography of the AF-2 surface.
Interaction fingerprint analysis revealed that all
the leads exhibited similar interactions with side
chain residue of Glu 353 when compared with
that of 4-OHT. Similarly, 4-OHT, lead 2 and lead
4 showed H-bond interactions with side chain
residue of Arg 394 in identical manner, indicating
antiestrogenic potential of proposed leads. E99
showed H-bond interaction with side chain
residue of Arg 394, H-bond interaction with
backbone residues of Leu 387 but lack of H-bond
interaction with Glu 353 in ligand binding domain
of hERα. In addition to the interactions with key
interacting residues, lead 1 and lead 2 showed
H-bond interaction with backbone residues of
Leu 387 in ligand binding domain of hERα
indicating more binding affinity of proposed leads
towards hERα than the best co-crystal ligand 4-
OHT.

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
478
Fig. 4. Docking interactions of lead 3 with hERα
Fig. 5. Docking interactions of lead 4 with hERα

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
479
Fig. 6. Docking interactions of E99 with hERα
Fig. 7. Evaluation of proposed leads with Enrichment curve

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
480
Table 4. Interaction fingerprint of the docked complexes of the best co-crystal ligand 4-OHT, proposed leads and E99 with hERα
Compound
Glu 353
Leu 387
Arg 394
Any contact
H- bond (Backbone)
H-bond (Sidechain)
Charged (Negative)
Charged (Positive)
Hydrophobic
Polar
Pi-Pi stacking
Salt Bridge
Any contact
H- bond (Backbone)
H-bond (Sidechain)
Charged (Negative)
Charged (Positive)
Hydrophobic
Polar
Pi-Pi stacking
Salt Bridge
Any contact
H- bond (Backbone)
H-bond (Sidechain)
Charged (Negative)
Charged (Positive)
Hydrophobic
Polar
Pi-Pi stacking
Salt Bridge
4-OHT
1
0
1
1
0
0
0
0
0
1
0
0
0
0
1
0
0
0
1
0
1
0
1
0
0
0
0
Lead 1
1
0
1
1
0
0
0
0
0
1
1
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
Lead 2
1
0
1
1
0
0
0
0
0
1
1
0
0
0
1
0
0
0
1
0
1
0
1
0
0
0
0
Lead 3
1
0
1
1
0
0
0
0
0
1
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
Lead 4
1
0
1
1
0
0
0
0
0
1
0
0
0
0
1
0
0
0
1
0
1
0
1
0
0
0
0
E99
1
0
0
1
0
0
0
0
0
1
1
0
0
0
1
0
0
0
1
0
1
0
1
0
0
0
0

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
481
3.6 Lead Validation
Four proposed leads and co-crystal ligand 4-
OHT were taken as a query to screen against the
internal library of 1000 decoys and actives
resulted from XP docking with hERα, yielded
60% of known actives which were within EF1%
of internal library that comprises both actives and
decoys (EF1% = 60). ROC metric corresponds to
the position of actives to the orderly ranked
compounds that are linearly arranged among the
internal library defined. Truchon and Bayly
considered ROC with ≥ 0.7 as a suitable
execution measuring value (where, ROC was
limited to 0-1) [25]. In the present study all the
known actives were retrieved with ROC of 0.98
relative to the hERα inhibitor ligands in the virtual
screening. BEDROC (α=20) metrics measures
the early recognition enrichment of actives
among the ranked compounds from the internal
database. BEDROC value of 0.94 is a beneficial
value that embodies the magnitude of early
recognition of actives from the ranked
compounds in the internal library. Deduced EF,
ROC and BEDROC values were 100 (EF1%),
0.98, 1.0 (α =20.0) respectively, explains that
four leads were efficacious and sufficient in
retrieving the active compounds. The enrichment
curve graphically represents quality of retrieved
actives which were ranked after comparing to
decoys in the internal library (Fig. 7).
4. CONCLUSION
Binding of estrogen such as estradiol to ERα
induces tumor growth in most ERα-
positive breast cancer cell lines, selective
estrogen receptor modulators prevent estrogen-
responsive breast cancers by targeting the ERα.
In present study, 3ERT crystal structure was
selected for docking and interaction fingerprint
analysis to design novel estrogen receptor
inhibitors. The analysis of hERα-4-OHT
complexes revealed key amino acids present in
the binding site of hERα that are important for
ligand binding. The best docked existing ligand
and co-crystal ligand were used for machine
learning based virtual screening against in-house
library to efficiently find potential lead molecules
among millions of compounds and rigid receptor
docking was performed with the generated library
of hERα inhibitors. Four leads were finally
obtained as the outcomes of the study with better
binding affinity in terms of XPG scores, good
structural properties with molecular contacts,
pharmacological properties than the existing
compounds and co-crystal ligand. Interaction
fingerprint analysis further confirmed that
proposed leads exhibited interaction pattern
similar to that of co-crystal ligand with increased
binding affinity and favorable orientation towards
hERα, so that critical binding sites were blocked
in turn to reduce the activity of hERα in estrogen-
dependent tumor growth. The obtained leads
from rigid receptor docking studies were
analyzed for ADME/T properties to propose the
leads. The proposed leads were also validated
and ranked better than the existing co-crystal
ligand and decoys in ROC metrics. As the
proposed ligands are novel and can be
synthesized in the lab and further they will be
subjected for evaluation of antiestrogenic activity
by in vitro. Hence these leads were proposed as
novel inhibitors against hERα.
DISCLAIMER
The products used for this research are
commonly and predominantly use products in our
area of research and country. There is absolutely
no conflict of interest between the authors and
producers of the products because we do not
intend to use these products as an avenue for
any litigation but for the advancement of
knowledge. Also, the research was not funded by
the producing company rather it was funded by
personal efforts of the authors.
CONSENT
It is not applicable.
ETHICAL APPROVAL
It is not applicable.
ACKNOWLEDGEMENTS
Authors are thankful to DST-CURIE-AI Centre,
Sri Padmavati Mahila Visvavidyalayam,Tirupati
for providing financial assistance for this
projectwork.
COMPETING INTERESTS
Authors have declared that no competing
interests exist.
REFERENCES
1. Sharma GN, Dave R, Sanadya J, Sharma
P, Sharma KK. Various types and
management of breast cancer: An

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
482
overview. J Adv Pharm Technol Res.
2010;1(2):109-126.
PMID: 22247839; PMCID: PMC3255438.
2. Anonymous. Cancer cases in India
estimated to be 13.9 lakh in 2020, may rise
to 15.7 lakh by 2025: Report. The times of
India; 2020.
Accessed 1 Aug 2021.
Available:https://m.timesofindia.com/india/c
ancer-cases-in-india-estimated-to-be-13-9-
lakh-in-2020-may-rise-to-15-7-lakh-by-
2025-
report/amp_articleshow/77628680.cms
3. Anonymous. Statistics of Breast
Cancer in India. Cytecare Cancer
Hospitals; 2021.
Accessed 15 Aug 2021.
Available:https://cytecare.com/blog/statistic
s-of-breast-cancer/
4. Anonymous. Breast Cancer Facts &
Figures 2019-2020. Atlanta: American
Cancer Society, Inc; 2019.
Accessed 15 Aug 2021.
Available:https://www.cancer.org/content/d
am/cancer-org/research/cancer-facts-and-
statistics/breast-cancer-facts-and-
figures/breast-cancer-facts-and-figures-
2019-2020.pdf
5. Anonymous. Common Breast Cancer
Types. Cancer Treatment Centers of
America; 2021.
Accessed 21 Sep 2021.
Available:https://www.cancercenter.com/ca
ncer-types/breast-cancer/types/common-
breast-cancer-types
6. Anonymous. Breast cancer types: What
your type means; 2020.
Accessed 20 Aug 2021.
Available:https://www.mayoclinic.org/disea
ses-conditions/breast-cancer/in-
depth/breast-cancer/art-20045654
7. Silberman A. ER-Positive Breast Cancer:
Prognosis, Life Expectancy, and More.
Healthline Media; 2020.
Accessed 21 Aug 2021.
Available:https://www.healthline.com/healt
h/breast-cancer/er-positive-prognosis-life-
expectancy
8. Jiang CF, Shi ZM, Li DM, Qian YC, Ren Y,
Bai XM, et al. Estrogen-induced miR-196a
elevation promotes tumor growth and
metastasis via targeting SPRED1 in breast
cancer. Mol Cancer. 2018;17:83.
DOI: 10.1186/s12943-018-0830-0
9. Puranik NV, Srivastava P, Bhatt G, Mary
DJSJ,LimayeAM, Sivaraman J. Determinat
ion and analysis of agonist and antagonist
potential of naturally occurring flavonoids
for estrogen receptor (ERα) by
various parameters and molecular
modelling approach. Sci
Rep. 2019;9:7450.
Available:https://doi.org/10.1038/s41598-
019-43768-5
10. Anonymous. Selective Estrogen Receptor
Modulators (SERMs). Breastcancer.org;
2020.
Accessed 19 Aug 2021.
Available:https://www.breastcancer.org/trea
tment/hormonal/serms
11. Ellis AJ, Hendrick VM, Williams R, Komm
BS. Selective estrogen receptor
modulators in clinical practice: a safety
overview. Expert Opin Drug Saf.
2015;14(6):921-34.
12. Mary AC. Breast Cancer. MSD Manual
consumer version; 2020.
Accessed 25 Aug 2021.
Available:https://www.merckmanuals.com/h
ome/women-s-health-issues/breast-
disorders/breast-cancer
13. Anonymous . Tamoxifen and Raloxifene
for Lowering Breast Cancer Risk.
American Cancer Society; 2019.
Accessed 29 Aug 2021.
Available:https://www.cancer.org/cancer/br
east-cancer/risk-and-prevention/tamoxifen-
and-raloxifene-for-breast-cancer-
prevention.html
14. Anonymous. Aromatase Inhibitors.
Breastcancer.org; 2020.
Accessed 25 Aug 2021.
Available:https://www.breastcancer.org/tre
atment/hormonal/aromatase_inhibitors
15. Sebastian R, Benjamin K. Machine
learning and AI-based approaches for
bioactive ligand discovery and GPCR-
ligand recognition. Methods. 2020;180:89-
110.
Available:https://doi.org/10.1016/j.ymeth.20
20.06.016
16. Meng XY, Zhang HX, Mezei M, Cui M.
Molecular docking: a powerful approach for
structure-based drug discovery. Curr
Comput Aided Drug Des. 2011;7(2):146-
157.
DOI: 10.2174/157340911795677602.
17. Choudhary MN, Connolly J. Artificial
Intelligence in Medicine Discovery:AI
in Virtual Screening. 32nd Irish Signals
and Systems Conference (ISSC). 2021;1-
6.
18. Rácz A, Bajusz D, Héberger K. Life
beyond the Tanimoto coefficient: similarity

Rajitha et al.; JPRI, 33(46A): 470-483, 2021; Article no.JPRI.75474
483
measures for interaction fingerprints. J
Cheminformatics. 2018;10(1):1-12.
19. Pradeep N, Munikumar M, Swargam S,
Hema K, Sudheer Kumar K,
Umamaheswari A. 197 Combinations of e-
pharmacophore modeling, multiple docking
strategies and molecular dynamic
simulations to discover of novel
antagonists of BACE1. J Biomol Struct
Dyn. 2015;1:129-130.
DOI: 10.1080/07391102.2015.1032834
20. Friesner RA, Banks JL, Murphy RB. Glide:
A new approach for rapid, accurate
docking and scoring: Method and
assessment of docking accuracy. J Med
Chem. 2004;47:1739–1749.
21. Pradeep N, Munikumar M, Swargam S,
Hema K, Sudheer Kumar K. Combinations
of e-pharmacophoremodeling, multiple
docking strategies and molecular dynamic
simulations to discover of novel
antagonists of BACE1. J Biomol Stuct Dyn.
2015;1:129-33.
22. Duch W, Swaminathan K, Meller J.
Artificial Intelligence Approaches for
Rational Drug Design and Discovery.
Current Pharmaceutical Design.
2007;13(14):1497–1508.
23. Sandeep S, Priyadarshini V, Pradhan D,
Munikumar M. Umamaheswari A. Docking
and molecular dynamics simulations
studies of human protein kinase catalytic
subunit alpha with antagonist. J Clin Sci
Res. 2012;1:15-23.
24. Hema K, Sandeep S, Pradeep N,
Umamaheswari A. In silico agonist for
human extracellular superoxide dismutase
SOD3. Online J Bioinform. 2016;17(1):29-
40.
25. Madhulatha NR, Pradeep N, Sandeep S,
Hema K, Chiranjeevi P, Kumar KS, et al.
E-Pharmacophore Model Assisted
Discovery of Novel Antagonists of nNOS.
Biochem Anal Biochem. 2017;6(1):1-9.
DOI: 10.4172/2161-1009.1000307
26. Deng Z, Chuaqui C, Singh J. Structural
Interaction Fingerprint (SIFt): A Novel
Method for Analyzing Three-Dimensional
Protein-Ligand Binding Interactions. J Med
Chem. 2004;47:337-344.
27. Shiau AK, Barstad D, Loria PM, Cheng L,
Kushner PJ, Agard DA, et al. The
structural basis of estrogen
receptor/coactivator recognition and the
antagonism of this interaction by
tamoxifen. Cell. 1998;95:927–937.
28. Thirupathi D, Bhargavi K, Ramasree D,
Aboubakr HA, Thirupathi M, Lavanya SP,
et al. Design of novel lead molecules
against RhoG protein as cancer target – a
computational study. J Biomol Struct Dyn.
2016;35(14):3119–3139.
29. Schrödinger LLC. Schrödinger Software
Release; 2015-2.
Accessed 6 Sep 2021.
Available:http://gohom.win/ManualHom/
Schrödinger/
Schrödinger_2012_docs/general/qikprop_p
rops.pdf
30. Brylinski M. Aromatic interactions at the
ligand-protein interface: Implications for the
development of docking scoring
functions. Chem Biol Drug Des. 2018;
91(2):380-390.
© 2021 Rajitha et al.; This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.
Peer-review history:
The peer review history for this paper can be accessed here:
https://www.sdiarticle4.com/review-history/75474