DOI 10.

1007/s11094-021-02441-w
Pharmaceutical Chemistry Journal, Vol. 55, No. 5, August, 2021 (Russian Original Vol. 55, No. 5, May, 2021)

STRUCTURE-BASED VIRTUAL SCREENING

AND MOLECULAR DYNAMICS SIMULATION
TO IDENTIFY POTENTIAL SARS-COV-2 SPIKE
RECEPTOR INHIBITORS FROM NATURAL
COMPOUND DATABASE

Arkadeep Sarkar,1,# Debanjan Sen,1,# Ashutosh Sharma,2

Ravi Kumar Muttineni,3 and Sudhan Debnath4,*

Original article submitted August 25, 2020.

The outbreak of respiratory disease, COVID-19 caused by SARS-CoV-2 has now been spread globally and the
number of new infections is rising every moment. There are no specific medications that are currently avail-
able to combat the disease. The spike receptor of SARS-CoV-2 facilitates the viral entry into a host cell and
initiation of infection. Targeting the viral entry at the initial stage has a better advantage than inhibiting it in
later stages of the viral life cycle. This study deals with identification of the potential natural molecule or its
derivatives from MolPort Databank as SARS-CoV-2 spike receptor inhibitors using structure-based virtual
screening followed by molecular dynamics simulation. On the basis of ADME properties, docking score,
MMGBSA binding energy, 150 ns molecular docking studies, and final molecular dynamics analysis, two nat-
ural compounds – 3 (MolPort-002-535-004) docking score –9.10 kcal mol-1 and 4 (MolPort-005-910-183)
docking score –8.5 kcal mol-1, are selected as potential in-silico spike receptor inhibitors. Both hits are com-
mercially available and can be further used for in-vitro and in-vivo studies. Findings of this study can facilitate
rational drug design against SARS-CoV-2 spike receptor.
Keywords: MolPort database; spike receptor; ADME filtration; molecular docking; molecular dynamics.

1. INTRODUCTION drome coronavirus (SARS-CoV) [2] and Middle East Respi-

ratory Syndrome coronavirus (MERS-CoV) [3] appeared in
The outburst of SARS-CoV-2 infection has created an years 2002 and 2012, respectively. The two viruses origi-
emergency situation globally. Starting from the epicenter of nated from animal reservoirs, had caused global epidemics
COVID-19, Wuhan city of China, the infection rapidly with fatality rate of which were 9.5 and 34%, respectively
spread worldwide. Coronavirus (CoVs) species are envel- [4]. The third pathogenic coronavirus namely SARS-CoV-2
oped positive-sense single strand-largest RNA viruses be- was identified to be the reason for a typical pneumonia infec-
longing to the Coronaviridae family of the order Nidovira- tion outbreak [5 – 9]. Due to the high dissemination capacity
les, which are divided into four genera (a, b, g, and d) [1]. In of SARS-CoV-2, it has created a pandemic situation to pub-
the last two decades, two highly pathogenic human corona- lic health throughout the world and potentially brings a chal-
viruses (HCoVs) namely, Severe Acute Respiratory Syn- lenge to the scientific community to control the spread of the
virus. According to Worldometer’s COVID-19 data, as of 16
1
Department of Pharmacy, BCDA College of Pharmacy & Technology, August 2020, about 21.6286 million confirmed cases of
Jessore Road South, Hridaypur, Kolkata, West Bengal, 700127 India. COVID-19 infection with 0.7691 million death cases were
2
School of Engineering and Sciences, Tecnologico de Monterrey, Campus
Queretaro, Av. Epigmenio Gonzalez, No. 500 Fracc., San Pablo, 76130 reported over 215 countries [10]. Due to the absence of any
Queretaro, Mexico. effective treatment and medication against SARS-CoV-2 in-
3
Immunocure Discovery Solution Pvt. ltd. IKP, Hydrabad, 500078 India. fection and the high dissemination rate of the infection
4
Department of Chemistry, M. B. B. College, Agartala, Tripura, 799004 In-
dia.
worldwide, prompted the WHO to declare the SARS-CoV-2
#
Equal contributions. outbreak as a global public health emergency of international
*
e-mail: [email protected] concern [11, 12]. This urgent situation is pressing the world

441
0091-150X/21/5505-0441 © 2021 Springer Science+Business Media, LLC
442 Arkadeep Sarkar et al.

scientific community to respond with the development of host body [22 – 24]. Therefore, the spike receptor binding
novel vaccine or small molecule therapeutics for domain (RBD) of SARS-CoV-2 virus can be a potential tar-
SARS-CoV-2 [13, 14]. In the emergency situation drug get for the treatment and development of new drugs against
repurposing may be a short-term and non-specific solution to COVID-19 viral pathogenesis.
treat COVID-19 patients [15], therefore the development of Li, et al. [25] identified various traditional herbal medi-
more target oriented natural inhibitors is the need of the hour. cines as potential ant-viral against SARS-CoV infection al-
Depending on the target, therapeutics against COVID-19 though the mechanism of anti-viral activities is not clearly
disease can be divided into two categories: one is acting on understood. There are evidences that plants also exhibit
the human immune system or host cells, and the other is on anti-viral activities against other coronavirus species [26].
SARS-CoV-2 life cycle. The therapeutics acting on the The mechanism of action is mainly through the inhibition of
coronavirus includes inhibiting the viral RNA through inter- the viral replication process and prevention of viral entry into
acting with the genetic material of the virus. Preventing the the host cell. Baicalin and baicalein are two natural flavones
viral replication process by interacting with its crucial life showing potential in vitro inhibitory activity against
cycle enzymes or by resisting the viral spike receptor binding SARS-CoV-2 main protease [27]. Baicalein is available from
to human ACE-2 or inhibiting the virus’s self-assembly pro- Oroxylum indicum, the fruit of which is used as vegetable in
cess through acting on some structural proteins of the virus North East India [26]. Similarly, docking study of seventeen
[16]. Depending upon the protein sequence alignment analy- organosulfur compounds available in garlic essential oil
ses of SARS-CoV-2, it has proven that, the genomic RNA of
showed that these compounds may exhibit good binding af-
SARS-CoV-2 contains 29,736 nucleotides. The envelope and
finity with human ACE2 and SARS-CoV-2 main protease.
nucleocapsid proteins of SARS-CoV-2 are two evolutionarily
The in-silico study suggests that the garlic oil may be an im-
conserved regions, having the sequence identities of 96% and
portant natural antivirus source that can prevent entry of
89.6%, respectively, compared to SARS-CoV [17]. Whereas,
SARS-CoV-2 into the human cell [18].
the spike protein of SARS-CoV-2 exhibited 77% sequence
In the present study, a total of 1,20,000 natural com-
similarities with SARS-CoV [12]. The spike is a
pounds were initially selected from the MolPort database
transmembrane glycoprotein that promotes the host cell at-
tachment and virus-cell membrane fusion during virus infec- and, depending upon their pharmacokinetic properties, about
tion. SARS-CoV-2 utilizes the homotrimeric glycoprotein of 50,000 of compounds were filtered through pkCSM open
each spike, which consists of two monomeric subunits (S1 source server. A virtual screening based molecular docking
sub-unit and S2 sub-unit), to bind with the host cell receptor. of selective ligands against the receptor binding domain
Recently, Lan, et al. [18] showed that SARS-CoV-2 viral (RBD) of SARS-CoV-2 was performed. Depending upon the
spike S1 sub-unit receptor binding domain (RBD) is able to binding interaction with binding threshold value > -8.00 kcal
utilize host receptor angiotensin-converting enzyme 2 mol-1, top 4 molecules were selected. To analyze the stability
(ACE2) as an entry receptor in ACE2-expressing cells, sug- of the protein-ligand complex and backbone, NVT and NPT
gesting potential drug targets for therapeutic development. In ensemble 50 ns molecular dynamics (MDs) simulations were
addition, the viral entry requires, the spike proteins priming performed for each molecule and the RMSD (root mean
through cellular proteases of the host cell, which promotes square deviation) and RMSF (root mean square fluctuation)
spike (S) protein cleavage at the S1/S2 site and facilitates the values of protein-ligand complex and backbone were re-
viral fusion with the host cell [19 – 22]. Past cryo-electron ported. Most of the works related to SARS-CoV-2 drug de-
microscopy investigations of the SARS-CoV-2 spike protein sign are re-purposing the existing drugs or in-silico identifi-
interaction with the host cell ACE2 receptor have indicated cation of natural compounds. But in our study, the selected
that receptor attachment initiates the separation of the viral molecules are natural products or natural product derivatives
S1 sub-unit with ACE2 receptor. This emphasizes the S2 which are commercially available. Therefore, scientific com-
sub-unit to form an increasingly steady post-fusion state that munity can easily explore their biological activity in wet lab-
is basic for membrane fusion [19]. Furthermore, cryo-EM oratory.
structure of the spike protein and biophysical assays reveal
that the SARS-CoV-2 spike binds with ACE2 in higher affin- 2. MATERIALS AND METHODS
ity than SARS-CoV [23]. The receptor binding domain
(RBD) of the S1 sub-unit stabilizes membrane attached S2 2.1. Data Set
sub-unit of the virus [18, 24]. A potential ACE2 inhibitor
arbidol is suggested for the treatment COVID-19 infection, The x-ray crystal structure of the recently discovered
but due to the low selectivity and a number of side effects, SARS-CoV-2 spike receptor (PDB ID: 6M0J, Resolution:
the ACE2 receptor inhibitor may not be the potential thera- 2.45Å) [19] was retrieved from RCSB Protein Data Bank
peutic against COVID-19 infection. It was also revealed in (www.rcsbpdb.org) in PDB format. The MolPort Database
recent studies that the virus can enter through the [https://www.molport.com] which contains 1.12 lakh com-
transmembrane cellular serine protease (TMPRSS2) receptor mercially available natural compounds and related deriva-
of the host cell and facilitates the viral pathogenesis in the tives were used for virtual screening.
Structure-Based Virtual Screening and Molecular Dynamics Simulation 443

2.2. Protein Preparation

The protein was imported in Auto Dock Tools 1.5.6 and
removed chain A, water molecules, and hetero atoms [28]
and saved it as pdb format. This E chain of protein was used
for Computed Atlas of Surface Topography of proteins
(CASTp 3.0) server [29] and identified the binding site
amino acid residues (338, 339, 342, 343, 367, 368, 371, 373,
and 374), volume and area of the active site pocket were
60.837 Å3 and 73.389 Å2, respectively. This server uses the
alpha shape method to determine the aforementioned topo-
graphic features for the input PDB file. The binding site in-
formation generated by CASTp 3.0 server was used to gener-
ate the grid box [30].
2.3. Preparation of Protein for Vina
The previously saved pdb format of the protein was im-
ported and polar hydrogens were added to the protein. Then
necessary charges like, Gasteiger and Kollman charges were
added. The protein was assigned to Auto Dock Tools 1.5.6
and was exported as PDBQT format. Depending upon the
CASTp 3.0 binding site information the grid box was pre-
pared and saved as a text file. The grid box dimension centre
was of X, Y and Z set to –36.022, 9.043 and 29.693, respec-
tively. Fig. 1. Virtual Screening workflow for identification of
SARS-CoV-2 spike receptor inhibitors.
2.4. Preparation of Ligands
A total of 120,000 natural compounds or natural com-
pound derivatives were retrieved from Mol-Port Database.
The drug-likeness property of the molecules depends upon ing the aforementioned script written program [32]. The
the absorption, distribution, metabolism, excretion and toxi- analysis of ligand binding contribution and the semi-empiri-
cological (ADMET) and pharmacodynamics properties. cal force field were performed in AutoDock Tool. The bind-
Based on the Lipinski’s rule of five, these molecules were fil- ing poses of the ligands were generated by using Lamarckian
tered, depending upon the absorption, distribution, metabo- Genetic Algorithm. Other docking parameters like, popula-
lism, excretion and toxicological properties. The ADMET tion size was set to 150, and maximum numbers of evalua-
properties of these compounds were predicted by open tions were set to 2500000. Maximum number of generations
source server pkCSM [31] and 50,000 numbers of hits were was chosen up to 27000, rate of gene mutation was set to
filtered out for further studies. These molecules were con- 0.02, rate of crossover was set to 0.8 and the exhaustiveness
verted to PDBQT format by using a script written program of was set to 8. The other parameters in Auto Dock Tool were
Samadani, et al. [32]. The force field was used during ligand set as default. Depending upon the standard-root-mean-devi-
preparation was Gaff force field [33] and by applying Ge- ation values and binding free energy, the results were clus-
netic algorithm, the best structural conformer was generated. tered. The resulted protein-ligand complex was observed by
The ligands were prepared using a steepest descent algorithm UCSF Chimera 1.13 software [35] and LigPlot [36]. Based
in which 1000 steps of energy minimization of the ligands upon the binding interaction and binding score above
were carried out. -8.00 kcal mol-1 top hits were selected for further analysis.
To determine the stability of these top four protein-ligand
2.5. Virtual Screening of Ligands with SARS-CoV-2 complexes at 50 ns NVT and NPT ensemble molecular dy-
Spike-Receptor Binding Domain namics were performed for each complex.
The structure-based virtual screening (VS) is an impor-
2.6. Molecular Dynamics Stimulation Study
tant approach to identify the potential in-silico inhibitors
against a particular target protein. The hits having good bind- The molecular dynamics (MD) simulations were carried
ing affinity towards a receptor are identified from the large out up to 50 ns to analyze the stability of the receptor-ligand
database. Structure-based virtual screenings of 50,000 mole- complexes of six selected ligands using Desmond software
cules with SARS-CoV-2 spike receptor binding domain was [37, 38]. Before simulation the system was solvated in a
performed using Autodock Vina [34]. The VS work flows of TIP3P water model and 0.15 M NaCl was incorporated to
the study are presented in Fig. 1. The VS was performed us- neutralize and impersonate the physiological condition. The
444 Arkadeep Sarkar et al.

simulation was run for 50 ns at 315.10 K temperature [39] sure bath [41]. The Particle Mesh Ewald method [42] was
and 1.01325 bar pressure. The periodic boundary conditions used with a cut-off distance of 9.0Å to determine the electro-
of grid box were enforced in all directions and prepared the static interaction. The root mean square deviation (RMSD)
simulation chamber by an auto calculated orthorhombic box of the protein-ligand complex and backbone were calculated
with buffer dimensions of 10 Å ´ 10 Å ´ 10 Å and NPT en-
with respect to the reference structure (Rref) by using the fol-
semble. Applying a steepest descent algorithm with 50,000
lowing equation:
steps maximum and OPLS-2005 force field, the energy
minimization and moderation of the position of the solvated 1
system of the protein-ligand complexes were carried away. é1 N 2ù 2
The Parrinello-Rahman coupling method was employed [40] RMSD( t ) = ê å m i ri ( t ) - riRf ú (1)
for the equilibration of both the NVT and NPT ensemble. In ëM i =1 û
case of NVT ensemble, a constant number of particles (N),
volume (V), temperature (T) and coupling constant 0.1 ps for where M = Simi and ri(t) describes the position of ith atom at
100 ps were maintained throughout the MD simulation time. time t, after Least Square (LS) fitting of a given molecule to
In case of NPT ensemble, a constant number of particles (N), the reference molecule. Root Mean Square Fluctuation
pressure (P), temperature (T) and equal coupling constant (RMSF) is defined as a measure of the fluctuation between
were maintained. The pressure during MD stimulations was the position of ith particle and reference position and deter-
maintained at 1 atm using the Martyna–Tobias Klein pres- mined by the following equation:

TABLE 1. Structures of Selected Hits with MolProt ID, Details of H-Bonding Interactions and Hydrophobic Interactions between Receptor
-1
Binding Domain Amino Acids of SARS-CoV-2 Spike Receptor and Their Interaction Distance and Binding Energy (kcal mol ).

Compound Structure with MolProt ID Interacting amino acid residue

1 Hydrogen bond: CYS-336 (2.94Å, 2.83Å), VAL-362
(3.03Å), ASP-364 (2.84Å), SER-373 (3.05Å)
Hydrophobic: LEU-335, PHE-342, ALA-363, VAL-367,
LEU-368, SER-371, PHE-374 TRP-436

MolPort-000-856-466, Binding energy: –9.1

2 Hydrogen bond: SER-371 (2.98Å), SER-373 (3.15Å,
3.24Å)
Hydrophobic: LEU-335, PHE-338, GLY-339, PHE-342,
ASN-343, ASP-364, LEU-368

MolPort-002-535-004, Binding energy: –8.9

3 Hydrogen bond: CYS-336 (3.10Å, 3.01Å), GLY-339
(3.09Å), VAL-362 (3.00Å), ASP-364 (3.10Å),
SER-373 (3.19Å)
Hydrophobic: LEU-335, PHE-338, PHE-342, ASN-343,
ALA-363, VAL-367, LEU-368, SER-371

MolPort-005-910-183, Binding energy: –8.7

4 Hydrogen bond: ASP-364 (2.82Å)
Hydrophobic: LEU-335, CYS-336, GLY-339, PHE-338.
PHE342, ASN-343, ALA-363, VAL-367, LEU-368,
SER-371, SER-373, SER-374

MolPort-002-225-685, Binding energy: –8.5

Structure-Based Virtual Screening and Molecular Dynamics Simulation 445

Fig. 2. LigPlot images showing hydrogen bonding and hydrophobic interactions: (1) ligand 1-6M0J interactions; (2) ligand 2-6M0J interac-
tions; (3) ligand 3 – 6M0J interactions; (4) ligand 4-6M0J interactions. Circled residues represent hydrophobic interaction with active site resi-
dues.

1
2.7. ADMET Predictions
é1 T 2ù 2
RMSD( t ) = ê å ri ( t j ) - riref ú (2) The absorption, distribution, metabolism, excretion and
êë T ti =1 úû
toxicity (ADMET) properties of the selected molecules was
predicted by the open source server pkCSM-pharmacoki-
where, T denotes the time over which one wants to calculate netics [31]. The absorption properties of these molecules
the average fluctuation and riref indicates the reference posi- were analyzed by using several physicochemical and phar-
tion of ith particle. macological parameters like, solubility, human colorectal
446 Arkadeep Sarkar et al.

Fig. 3. Surface diagrams of SARS-CoV-2 spike receptor and ligands 1, 2, 3 and 4 (shown in yellow stick).

carcinogenicity, human intestinal absorption (HIA%), skin property, Ames test results, carcinogenicity bioassay analysis
permeability, in-vivo P-glycoprotein inhibition parameter. on mice.
By calculating various parameters like, Cytochrome P450
2C19, 2C9, 2D6 and 3A4 enzyme inhibition guideline, meta- 2.8. Binding Free Energy Calculations
bolic properties of the selected hits were characterized. The The average binding free energies (DG bind) were calcu-
drug-likeness properties of these molecules were evaluated lated from the molecular dynamics trajectory. It infers impor-
using the Lipiniski’s Rule of Five calculation, blood-brain tant information about the affinity of the ligands towards re-
barrier permeability and central nervous system permeability ceptor. MM-GBSA (Generalized-Born Surface Area) meth-
property. The toxicity assessment of the molecules were car- ods based binding energy was calculated using
ried away by evaluating parameters like acute algae toxicity thermal mmpbsa.py script program. Every frames of 50ns

TABLE 2. Eestimated Binding Energy, Binding Affinity, and Average RMSD and RMSF of Ligand-Protein Complexes for 50-ns Molecular
Dynamics Simulations
MMGBSA (Binding
System Avg RMSD (Å) Avg RMSF (Å) Avg Rg (Å) SASA of protein (Å2) PSA (Å2)
energy) kcal mol-1
Apo protein 2.90 ± 0.42 1.47 ± 0.91 – – –
L1-6m0j 2.97 ± 0.67 1.79 ± 1.13 4.82 ± 0.11 207.32 ± 51.05 182.26 ± 12.96 –52.1467 ± 0.83
L2-6m0j 3.25 ± 0.65 1.27 ± 0.86 4.12 ± 0.39 315.25 ± 48.21 104.25 ± 7.42 –48.8684 ± 1.21
L3-6m0j 2.61 ± 0.49 1.36 ± 1.07 5.08 ± 0.13 207.99 ± 30.20 200.88 ± 5.07 –59.6584 ± 0.62
L4-6m0j 1.96 ± 0.321 1.01 ± 0.61 4.95 ± 0.07 166.92 ± 31.07 159.06 ± 2.89 –55.6087 ± 0.41
Structure-Based Virtual Screening and Molecular Dynamics Simulation 447

Fig. 4. RMSD of apo-protein and protein-ligand complex in 50 ns molecular dynamics simulations. Different colors in the figure refer to apo
protein (black), ligand 1-6M0J complex (red), ligand 2-6M0J complex (green), ligand 3-6M0J complex (blue), and ligand 4-6M0J complex
(yellow).

molecular dynamics trajectory were used as an input for this (MolPort-000-856-466), 2 (MolPort-002-535-004), 3
script program [46] (MolPort-005-910-183), and 4 (MolPort-002-225-685) hav-
ing nice binding affinity toward the active site of spike recep-
3. RESULTS AND DISCUSSION tor. The docking score of compounds 1, 2, 3 and 4 was –9.1,
–8.9, –8.7 and –8.5 kcal mol-1, respectively, as shown in Ta-
SARS-CoV-2 spike receptor binding domain plays a cru- ble 1. Among the selected compounds, the binding affinity of
cial role in viral entry into the host cell, as it directly binds hit compound 1 was highest and that of 4 was the lowest.
with the ACE2 receptor on the host cell [19]. Recently dis- The 2D ligand interactions of these hits were plotted by
covered x-ray crystallographic structure of the SARS-CoV-2 LigPlot software [36] as shown in Fig. 2. It was found that
spike receptor created possibility to develop drugs against the binding energy of compound 1 was –9.1 kcal mol-1 and
the SARS-CoV-2 spike receptor binding domain to inhibit the interacting active site amino acid residues of the spike re-
viral entry into the host cell. In step 1, the database was fil- ceptor through hydrogen bonding were CYS-336 (2.94Å,
tered to 50,000 compounds depending upon the ADMET
properties through pkCSM open source server and filtered
compounds were used for structure-based VS studies. In step
2, the VS of ADME filtered hits were performed with the TABLE 3. Average RMSD of Protein-Ligand Complexes of Com-
binding site of the SARS-CoV-2 spike-RBD using Autodock pounds 3 and 4 for 150-ns Molecular Dynamics Simulations
Vina [34]. The hits with docking score greater than -8.0 were Name of system Average RMSD (Å)
further used for MD simulations using Desmond package Apo-protein 2.90 ± 0.36
[38]. The interacting active site amino acid residues with in- 3-Complex 2.82 ± 0.35
teraction distance of selected hits are presented in Table 1.
4-Complex 2.16 ± 0.33
The docking results indicated four hits: 1
448 Arkadeep Sarkar et al.

Fig. 5. RMSF of apo-protein and protein-ligand complexes during 50 ns molecular dynamics simulations. Different colors in the figure refer to
apo-protein (deep blue), ligand 1-protein complex (red), ligand 2-protein complex (yellow), ligand 3-protein complex (green), and ligand 4-pro-
tein complex (violet).

2.83Å), VAL-362 (3.03Å), ASP-364 (2.84Å), and SER-373 Compound 2 interacts through three H-bonds and seven hy-
(3.05Å). This compound showed hydrophobic interactions drophobic interactions. Six H-bonds and eight hydrophobic
with amino acid residues LEU-335, TRP-436, PHE-342, interactions are shown by compound 3. Similarly, one
ALA-363, VAL-367, LEU-368, SER-371, and PHE-374 of H-bond and twelve hydrophobic interactions are shown by
the spike receptor binding domain. The binding energy of compound 4. Except compound 4, all other hits have fair
compound 2 is –8.9 kcal mol-1 and it interacts with spike re- number of H-bonds and all the hits exhibit good number of
ceptor binding domain amino acid residues through hydro- hydrophobic interactions. All the hits are deeply inserted into
gen bonding were SER-371 (2.98Å), SER-373 (3.15Å, the active site of receptor and the surface diagram of
3.24Å). The hydrophobic interactions of this compound with SARS-CoV-2 spike receptor with ligands 1, 2, 3 and 4 are
residues were LEU-335, PHE-338, GLY-339, PHE-342, shown in Fig. 3.
ASN-343, ASP-364, and LEU-368. The binding energy of In this work, 50-ns MD simulations were performed for
compound 3 was –8.7 kcal mol-1 and it interacted with spike apo-protein and selected ligand-spike receptor (6M0J) com-
receptor binding domain residues through hydrogen bonding plexes to analyze stability the of receptor-ligand complex.
were CYS-336 (3.10Å, 3.01Å), GLY-339 (3.09Å), VAL-362 The calculated root mean square deviation (RMSD), from
(3.00Å), ASP-364 (3.10Å), SER-373 (3.19Å). This com- first to last stimulated trajectories provided the structural and
pound exhibited hydrophobic interactions with LEU-335, conformational information of the protein-ligand complexes.
PHE-338, PHE-342, ASN-343, ALA-363, VAL-367, The average RMSD of apo-protein backbone was ~2.90Å
LEU-368, and SER-371. The binding energy of compound 4 (Table 2) which was stable throughout MD simulations run.
was –8.5 kcal mol-1 and it interacted with spike receptor Average RMSD of the spike receptor (6M0J)-ligand com-
binding domain through hydrogen bonding with ASP-364 plexes of 1, 2, 3 and 4 were ~2.97Å, ~3.25Å, ~2.61Å and
(2.82Å) and hydrophobic interactions with LEU-335, ~1.96Å and ~2.35Å, respectively, are shown in Fig. 4 and
CYS-336, GLY-339, PHE-338, PHE342, ASN-343, Table 2. The RMSD of receptor-ligand complexes of mole-
ALA-363, VAL-367, LEU-368, SER-371, SER-373, and cules 3 and 4 were compared with apo-protein, i.e. after
SER-374. binding in the active site these ligands showed a small
All structures of selected hits with their binding energies, change in the receptor backbone during the simulation time.
interacting amino acid residues, and MolPort ID, are shown The RMSD of ~2.94Å was observed after the ~67 ns in the
in Table 2. Compound 1 showed four H-bonds and eight hy- trajectory of molecule 1-6m0j complex system probably due
drophobic interactions with the active site of spike receptor. the presence of higher degree of rotatable bonds or unable to
Structure-Based Virtual Screening and Molecular Dynamics Simulation 449

Fig. 6. RMSD of apo-protein and protein-ligand complex in 150 ns molecular dynamics simulations. Different colors in the figure refer to apo
protein (black), ligand 3-6M0J complex (blue), and ligand 4-6M0J complex (red).

attain the structural flexibility inside the binding cavity of the the apo-protein, although there was a small deviation ob-
spike protein. For the rest of the stimulation time the com- served in the residue range 356 – 364 amino acid (AA) for
plex system exhibit an average RMSD of ~3.16Å. The mole- ligand 3 (Fig. 5). Lower the values of RMSF indicate that the
cule 2-6m0j complex system exhibited an undesirable devia- reduced random motions and minimum fluctuations of pro-
tion of ~2.99Å in RMSD at ~10.5 ns and for the entire stimu- tein backbone and side chain during the simulation run.
lation the average RMSD was observed ~3.55Å. There was a Compound 1 showed maximum hydrophobic interactions
deviation observed at ~29.3 ns with a RMSF of ~3.17Å for with PHE-342, hydrogen bonding interaction with ASP-364
molecule 3-6m0j complex system. The highest fluctuation of throughout the simulation time. The active site residues
RMSD (~3.59Å) was observed at ~31.3 ns in the entire tra- ALA-372, PHE-374, and TRP-436 showed hydrophobic in-
jectory of molecule 3 – 6m0j complex system. But after the teractions with the compound and ASP-364, ASN-370,
fluctuation, the system exhibited complete stability with the SER-371, ALA-372 interacted with the ligand through wa-
spike protein with an average RMSD of ~3.10Å. Molecule 4 ter-bridge (Fig. S1). Compound 2 showed nice hydrophobic
exhibited better stability with the protein with an average interactions with PHE-342, VAL-367, PHE-374, and
RMSD of ~1.96Å. TRP-436 and small fraction of hydrogen bonding interaction
The root mean square fluctuation (RMSF) values are with SER-373, TRP-436, ASN-437 and small fraction of wa-
useful for identification of any changes in backbone atoms as ter bridge interactions with ASN-343, SER-373, ASN-440,
well as side chain atoms of protein. The average RMSF val- and ARG-509 (Fig. S2). Compound 3 showed nice hydrogen
ues for ligand 1, 2, 3 and 4 complex system were observed bonding interaction with ASP-364 and small hydrogen bond-
~1.79Å, ~1.27Å, ~1.36Å and~1.01Å respectively, and the ing interaction with PHE-338. Other than this it also showed
average RMSF of apo-protein was found to be ~1.47Å. The hydrophobic interactions with PHE-342, PHE-374, TRP-436
RMSF values of all the selected hits were almost lower than and small water bridge interactions with ASP-364 and
450 Arkadeep Sarkar et al.

Fig. S1. Histogram of spike receptor residues interacting with MolPort-000-856-466 (interaction fractions) throughout 50 ns simulation time:
(A) MolPort-000-856-466 atom interactions with the spike receptor residues; (B) bar graph showing receptor 6M0J-compound 1, H-bond inter-
actions (green color), hydrophobic interactions (gray violet color), and water bridges (blue color).

TABLE 4. ADME Properties of Selected Hits

Parameter Ligand 1 Ligand 2 Ligand 3 Ligand 4
Molecular weight 397.427 474.55 373.405 368.385
Hydrogen bond donors (HBD) 1 2 2 1
Hydrogen bond acceptors (HBA) 5 5 5 5
Number of rotable bonds 3 9 5 4
logP 3.33 4.75 2.33 4.15
Absorption and distribution
#1
BBB+ Yes Yes Yes Yes
#2
HIA(%) 71.55 91.06 62.28 98.8
Aqueous solubility (LogS) –4.124 –5.481 –3.759 –4.109
Skin permeability (log Kp) –2.75 –2.73 –2.72 –2.73
Metabolism
CYP 2D6 inhibitory Promiscuity No No No No
Excretion
Total Clearance (log ml/min/kg) 0.716 1.026 0.581 0.625
Toxicity
Acute oral toxicity (mol/kg) 2.538 2.239 2.856 2.559
#1
BBB+: blood-brain barrier permeability; #2HIA(%): human intestinal absorption
Structure-Based Virtual Screening and Molecular Dynamics Simulation 451

Fig. S2. Histogram of spike receptor residues interacting with MolPort-002-535-004 (interaction fractions) throughout 50 ns simulation time:
(A) detailed MolPort-002-535-004 atom interactions with the spike receptor residues; (B) bar graph showing receptor 6M0J-compound 2,
H-bond interactions (green color), hydrophobic interactions (gray violet color), and water bridges (blue color).

Fig. S3. Histogram of spike receptor residues interacting with MolPort-005-910-183 (interaction fractions) throughout 50 ns simulation time:
(A) detailed MolPort-005-910-183 atom interactions with the spike receptor residues; (B) The bar graph showing receptor 6M0J-compound 3,
H-bond interactions (green color), hydrophobic interactions (gray violet color), and water bridges (blue color).
452 Arkadeep Sarkar et al.

Fig. S4. Histogram of spike receptor residues interacting with MolPort-002-225-685 (interaction fractions) throughout 50 ns simulation time:
(A) detailed MolPort-002-225-685 atom interactions with the spike receptor residues; (B) bar graph showing receptor 6M0J-compound 4,
H-bond interactions (green color), hydrophobic interactions (gray violet color), and water bridges (blue color).

SER-373 (Fig. S3). Compound 4 interacted with active site rier. The standard range of logBB value to permeate through
residues of spike receptor were ASP-364 through H-bonding, the blood-brain barrier lies between the range of 0.3 to –1
PHE-342, VAL-367 through hydrophobic, and ASP-364 and [43]. The molecules 1, 2, 3 and 4 exhibited log BB values
SER-373 through water bridge. The interaction fractions and –0.497, –0.485, –0.79 and –0.565, respectively (Table 4).
type of interactions throughout the simulation time are The maximum human intestinal absorption (HIA%) was ob-
shown in (Fig. S4). Compounds 3 and 4 showed stable pro- served for molecule 4 (HIA%- 98.8) and the others mole-
tein-ligand complex during the 50 ns molecular dynamics cules 1, 2 and 3 exhibited a HIA% of 71.55%, 91.06% and
simulation. Therefore, the two compounds were further stud- 62.28%, respectively (Table 4). The aqueous solubility plays
ied in 150-ns MD simulations. The average RMSD values of the important role in drug oral activity. Except molecule 3,
apo-protein, 3-spike receptor complex and 4-spike receptor all the molecules were found to be insoluble in water. The
complex were ~2.90,~ 2.82 and ~2.16Å, respectively, for standard range of LogS for aqueous solubility indicate that
150-ns simulation time (Fig. 6). The average RMSD of com- LogS >0 is highly soluble; molecules of Logs value 0 to –2
pound 3 and compound 4 are lower than the average RMSD are soluble; molecules with LogS value –2 to –4 are slightly
of apo-protein. Therefore, compound 3 and 4 showed stable soluble and LogS < –4 indicates insolubility of the molecule
binding with spike receptor (Table 3). The binding energy of in aqueous phase [44]. The LogS value of molecules 1, 2 and
compound 1-, 2-, 3-, and 4-spike receptor complex predicted 4 were found to be –4.124, –5.481 and –4.109, respectively
by prime are –52.1467, –48.8684, –59.6584, and (Table 4). Whereas, the molecule 3 exhibited moderate solu-
–55.6087 kcal mol-1. The binding energy values also support bility in water with a LogS value of –3.759 (Table 4).All the
the superiority of compound 3 and 4 as SARS-CoV-2 spike compounds have good skin permeability (logkp < 2.5) with
receptor inhibitors. the logkp value –2.75, –2.73, –2.72 and –2.73 for molecules
The summary of the ADMET properties of the selected 1, 2, 3 and 4, respectively (Table 3). The above observations
molecules are presented in Table 4. The molecular weights of indicated that the molecules have good absorption properties.
selected hits 1, 2, 3 and 4 were found to be 397.427, 474.55, The Cytochrome P450 (CYP 2D6) is an important enzyme
373.405, and 368.385 g mol-1, respectively. All molecules and plays the crucial role for the metabolism of more than 25
have a probability to permeate through the blood brain bar- % of drugs. None of these selected molecules exhibited any
Structure-Based Virtual Screening and Molecular Dynamics Simulation 453

inhibitory effect on this enzyme and indicated acceptable 6. A. Guihot, et al., Front. Immunol., 11, 1662 (2020).
metabolic properties. The predicted acute oral toxicity 7. W. Tai et al, Cell. Mol. Immunol., 17, 613 – 620 (2020).
8. C. Liu, Q. Zhou, et al., ACS Cent. Sci., 6, 315 – 331 (2020).
(LD50) value of the selected compounds were found to be
9. S. Xiu and A. Dick, J. Med. Chem., 63, 12256 (2020).
2.538 mol kg-1, 2.239 mol kg-1, 2.856 mol kg-1, 2.559 mol 10. https: // www.worldometers.info / coronavirus /
kg-1and 2.519 mol kg-1 (Table 4) for molecules 1, 2, 3 and 4, https://www.worldometers.info/coronavirus/
respectively. The results indicated a moderate toxicity of hit 11. S. A. Lauer, et al., Ann. Intern. Med., 172, 577 – 582 (2020).
molecules as per the standard value of LD50 (1 mol kg-1). 12. T. T. Y. Lam, et al., Nature, 583, 282 – 285 (2020).
The drug-likeness properties of any molecule depends 13. Y. Yang, et al., Virol. J., 17, 117 (2020).
upon various physicochemical properties like numbers of hy- 14. S.F. Ahmed, A. A. Quadeer, and M. R. McKay, Immunol. Stud.
Viruses, 12(3), 254 (2020).
drogen bond donors (HBD), numbers of hydrogen bond ac-
15. A. Savarino, J. Clin. Virol, 34, 170 – 178 (2005).
ceptors (HBA), molecular weight, number of rotatable bonds 16. M. Hoffmann, et al., Cell, 181, 271 – 280 (2020).
and logP value of that molecule (Lipiniski’s Rule of Five) 17. Y. Zhou, et al., Cell Discov., 6, 14 (2020).
[45]. For suitable drug like molecule, the HBD and HBA 18. J. Lan, et al., Nature, 581, 215 – 220 (2020).
must be in the range of 5 and 10, respectively. The molecular 19. B. T. P. Thuy, et al., ACS Omega, 5, 8312 – 8320 (2020).
weight must be < 500 gmol-1 and logP value should be less 20. J. Shang, et al., Nature, 581, 221 – 224 (2020).
than 5 [45]. In the present study, all the molecules met with 21. M. Bhattacharya, et al., J. Med. Virol., 92, 618 – 631 (2020).
the standard criteria for drug-likeness properties. For the 22. C. A. Devaux, J. M. Rolain, and D. Raoult, J. Microbiol.
Immunol. Infect. (2020).
molecule 1, 2, 3 and 4 the HBD was observed 1, 2, 2 and 1, 23. C. M. Yang, et al., Antiviral Res., 67, 24 – 30 (2005).
the number of HBA was estimated 5 for all these molecules 24. I. Astuti, Clin. Res. Rev., 14(4), 407 – 412 (2020).
and logP estimated values were found to be 3.33, 4.75, 2.33 25. S. Y. Li, et al., Antiviral Res., 67, 18 – 23 (2005).
and 4.15, respectively (Table 4). The results of physicoche- 26. B. Dinda, I. Silsarma, M. Dinda, and P. Rudrapaul, J. Ethno-
mical and pharmacological ADMET parameters of the se- pharmacol., 161, 255 – 278 (2015).
lected molecules indicated that, all the potential 27. H. Liu, et al., BioRxiv (2020).
SARS-CoV-2 spike protein inhibitors have suitable drug-li- 28. R. Huey and G. M. Morris, J. Scripps Res. Inst., 54 – 56 (2008).
29. G. E. Rhizobiu, Nucleic Acids Res., 1, 13 – 14(2013).
keness properties and good ADMET properties. Further 30. M. Sivaramakrishnan, et al., Inform. Med. Unlocked, 19,
in vivo and ex vivo study may be carried out to justify the 100331 (2020).
predicted properties of the molecule and to observe the po- 31. D. E. V. Pires, T. L. Blundell, and D. B. Ascher, J. Med. Chem.,
tentiality of these molecules against SARS-CoV-2 spike pro- 58, 4066 – 4072 (2015).
tein. 32. A. Samdani and U. Vetrivel, Comput. Biol. Chem., 74, 39 – 48
Based on the ADME properties, docking score and sta- (2018).
bility of protein-ligand complex during MD study, 33. K. G. Sprenger, V. W. Jaeger, and J. Pfaendtner, J. Phys. Chem.
B, 119, 5882 – 5895 (2015).
MMGBSA binding energy compound 3 and 4 were selected 34. O. Trott and A. J. Olson, J. Comput. Chem., 31 (2009)
as potential spike receptor inhibitors. These two compounds 35. E. F. Pettersen, et al., J. Comput. Chem., 25, 1605 –
showed stable protein-ligand complex throughout the 150 ns 1612(2004).
simulation time. All the molecules are commercially avail- 36. W. AC, L. RA, T. JM, Protein Eng., 8 (1995).
able and these two molecules may be used for further in vitro 37. M.I. Choudhary, M. Shaikh, A. Tul-Wahab, and A. Ur-Rahman,
and in vivo studies. PLoS One, 15(7), e0235030 (2020).
38. K. J. Bowers, et al., J. Inst. Electr. Electron. Eng. (IEEE), 5,
43 – 43 (2007).
CONFLICT OF INTEREST 39. D. J. Evans and B. L. Holian, J. Chem. Phys., 83, 4069 – 4074
(1985).
The authors declare that they have no conflict of interest. 40. P. Tangney, J. Chem. Phys., 124, 044111 (2006).
41. J. Li, A. H. W. Ngan, and P. Gumbsch, Acta Mater., 51,
5711 – 5742 (2003).
REFERENCES 42. T. Darden, L. Perera, L. Li, and L. Pedersen, Structure, 7,
55 – 60 (1999).
1. A. B. Durojaiye, J. R. D. Clarke, G. A. Stamatiades, and C. 43. S. Vilar, M. Chakrabarti, and S. Costanzi, J. Mol. Graph.
Wang, J. Biomol. Struct. Dyn., 2020, 1 – 8 (2020). Model., 28, 899 – 903 (2010).
2. Y. Zhou, et al., Antiviral Res., 116, 76–84(2015). 44. J. L. McDonagh, N. Nath, L. De Ferrari, T. Van Mourik, et al., J.
3. M. Prajapat, et al., Indian J. Pharmacol., 52, 56 – 65(2020). Chem. Inf. Model., 54, 844 – 856 (2014).
4. D. Schoeman and B. C. Fielding, Virol. J., 16(1) 1 – 22 (2019). 45. C. A. Lipinski, Adv. Drug Deliv. Rev., 101, 34 – 41 (2016).
5. N. H. Abd Ellah, et al., Nanomedicine, 15(21), (2020). 46. D. K. Yadav, et al. Front Chem., 8, 1 – 14 (2020)