Structure-guided pharmacophore based virtual screening, docking, and molecular dynamics to discover repurposed drugs as novel inhibitors against endoribonuclease Nsp15 of SARS-CoV-2

Abstract COVID-19 (Corona Virus Disease of 2019) caused by the novel ‘Severe Acute Respiratory Syndrome Coronavirus-2’ (SARS-CoV-2) has wreaked havoc on human health and the global economy. As a result, for new medication development, it's critical to investigate possible therapeutic targets against the novel virus. ‘Non-structural protein 15’ (Nsp15) endonuclease is one of the crucial targets which helps in the replication of virus and virulence in the host immune system. Here, in the current study, we developed the structure-based pharmacophore model based on Nsp15-UMP interactions and virtually screened several databases against the selected model. To validate the screening process, we docked the top hits obtained after secondary filtering (Lipinski’s rule of five, ADMET & Topkat) followed by 100 ns molecular dynamics (MD) simulations. Next, to revalidate the MD simulation studies, we have calculated the binding free energy of each complex using the MM-PBSA procedure. The discovered repurposed drugs can aid the rational design of novel inhibitors for Nsp15 of the SARS-CoV-2 enzyme and may be considered for immediate drug development. Graphical Abstract Communicated by Ramaswamy H. Sarma


Introduction
Corona Virus Disease of 2019, commonly referred to as COVID-19, is an infectious and respiratory disease caused by SARS-CoV-2 ('Severe acute respiratory syndrome coronavirus 2'), as identified by the World Health Organization. (Lai et al., 2020). It was initially recorded in China's Wuhan City (capital of Hubei Province) around the beginning of December 2019. Before the Wuhan government implemented a lockdown of the city, it spread rapidly in the other cities of China, and now the lethal virus grasped the whole world including India (Şahin et al., 2020). SARS-CoV-2 is a positive sense single-stranded virus that represents the 'Coronaviridae' family which is the largest of RNA viruses and is known for illness of the respiratory tract in humans similar to the Middle East respiratory syndrome (MERS) & Severe acute respiratory syndrome (SARS). SARS, MERS and SARS-CoV-2 have been registered under b-coronavirus genus (M. Pal et al., n.d.). This virus is closely related to SARS-CoV as it is highly homologous to it with more than 79% sequence identity but distantly related to MERS-CoV with only 50% homology . It contains four structural proteins: (i) Spike (S), (ii) Nucleocapsid (N), (iii) Matrix (M) and (iv) Envelope (E) and non-structural protein like RdRp (nsp12) and proteases (nsp3 and nsp5). Nsp3 and Nsp5 cleave the large polyproteins (pp1a and pp1ab) which were translated from replicase genes (rep1a and rep1b) resulting in 16 viral Nsps that congregate into replicase complex and have various enzymatic activities (Prajapat et al., 2020). From the various NSPs, Nsp15, a hexamer forming protein is a crucial member of the nidoviral RNA uridylatespecific endoribonuclease (NendoU) family. It has a C-terminal catalytic domain ( Figure 1A-C) that performs a variety of biological roles such as RNA endonuclease activity, producing 2 0 -3 0 cyclic phosphodiester termini. Nsp15 is conserved across nidoviruses but absent from other RNA viruses in viruses, making it a viable target for antiviral treatments (Kim et al., 2020).
Nsp15 help in the virus replication that may be by interference with host immune response (New Coronavirus Protein Reveals Drug Target, n.d.). Another study suggested that Nsp15 is responsible for the deterioration of viral RNA to mask it from the host immune system (Chandra et al., 2021). Nsp15 inhibition has been discovered to limit replication of the virus more efficiently than any other covid target (Batool et al., 2021). Previous studies also illustrated that a single nucleotide mutation of Nsp15 knocks down its endonucleolytic activity whereas deletion of Nsp15 remarkably decreases viral replication (Gao et al., 2021;Ivanov et al., 2004;Kang et al., 2007). Nsp15 in SARS-CoV-2 differentiates from other viruses, as it is observed as a genetic marker among Nidovirales order (Chandra et al., 2021). Nsp15 protein was also studied during the SARS outbreak (in the year 2003) for new drug development. All such programs failed then due to subsiding of the SARS epidemic at that time and the inhibitors never developed into drugs. These inhibitors have now been tested for COVID-19's Nsp15 protein (Ortiz-Alcantara et al., 2010). All the research institutions, pharmaceutical companies, and government agencies across the world have taken the challenge to discover new therapeutic agents against novel coronavirus (SARS-CoV-2).
Drug repurposing and in-silico drug design is the most widely used procedure for finding new targets for already used drugs or clinically/investigational drugs. In the current scenario, this strategy would be the prominent option for the identification of novel therapeutic agents for the SARS-CoV-2 virus considering the advantage of this method as repurposed drugs have already passed the pre-clinical and clinical trials. As the coronavirus cases are still escalating day by day, this method saves a lot of time and money (Sahoo et al., 2021). Currently, patients are being administered different repurposed drugs like Flavipiravir, Remdesivir, and Hydroxychloroquine to reduce the viral load but these drugs are not able to decrease the mortality rate among hospitalized COVID-19 patients (Michele et al., 2020). Therefore, there is a need to identify new inhibitors of other target proteins of SARS-CoV-2 including Nsp15.
To identify new, selective, and potential inhibitors against Nsp15 and obtain more effective drug-like candidates with low toxicity, many approaches were used via computer-aided drug design. (3, phenyl]butanamide) is a promising drug-like molecule reported by 'Kamal and co-workers' against Nsp15 by virtual screening and molecular dynamics approach (Batool et al., 2021). Ataul Islam et al. identified potential Nsp15 modulators using targeted docking and MD studies against an antiviral library (Savale et al., 2021). Voorhis's group reported antiviral activity of Exebryl-1 (1-20 mM) using a high-throughput assay in which initial hits were pan-inhibitors (Choi et al., 2021). Alazmi's group attempted to screen natural compounds via computational biology and proposed two molecules 95372568 and 1776037 as potential hits against Nsp15 (Motwalli & Alazmi, 2021). Nam-Chul Ha's group examined the antiviral activity of Epigallocatechin gallate using reduction neutralization tests with SARS-CoV2 strain after performing the in vitro screening of natural compounds (Hong et al., 2021) (Table ST 1).
In this study, we explored the uridine-5 0 -monophosphate binding site of Nsp15 ( Figure 1D,E) and constructed the structure-based pharmacophore model based on crucial interactions required for Nsp15 activity and report a multistep screening method consisting of database building (DrugBank compounds, nucleotide analogs, bioactive compounds, and antiviral compounds), pharmacophore modelbased high throughput screening (HTVS), molecular docking, filtering of hit-candidates and molecular dynamics (MD) to get novel Nsp15 inhibitors. The top ten hits were obtained through virtual screening followed by molecular docking and molecular dynamics studies. These new molecules may be further explored for their antiviral properties. Few of the molecules discovered through drug bank database search can directly be repurposed against NSP-15 ( Figure 1F).

Data collection & building of chemical database
The tertiary structure of Nsp15 Endoribonuclease-UMP(Uridine-5 0 -Monophosphate) complex from SARS-CoV-2 with x-ray resolution 1.82 Å was downloaded from the Protein data bank (https://www.rcsb.org), with PDB ID: 6WLC (Kim et al., 2021). The crystal structure of Nsp15 was further prepared through the protein preparation module (Baby et al., 2016) of Biovia Discovery Studio 2020. Next, we retrieved the DrugBank library (11172 compounds), Antiviral library-L7000 (488 compounds), Nucleotide analogue library-L7200 (156 compounds), and Bioactive compound Library-L1700 (7345 compounds) of Shelleckchem. All the compounds were downloaded in .sdf format and were incorporated in Biovia DS 2020 using the Build Database module. (Singh & Srivastava, 2014) These compounds were further prepared and energy minimized (4000 steps of conjugate gradient after 4000 steps of steepest descent) to pass the criteria of RMS gradient of 0.001 by using ligand preparation and energy minimization protocol (Jamal et al., 2012) of Biovia DS 2020.

Structure-based pharmacophore modeling
Receptor-ligand Pharmacophore Generation (RLPG) protocol was utilized to build a structure-based pharmacophore model based on the tertiary structure of SARS-CoV-2 Nsp15 complexed with Uridine-5 0 -Monophosphate (PDB ID:6WLC) (Kim et al., 2021;Kurczab & Bojarski, 2013). Before hypotheses generation the binding site of NSP15 was validated by redocking the co-crystallized ligand 'UMP' (Diallo et al., 2021) into a defined binding site and the ligand rmsd was calculated for the output conformations. Two pharmacophore hypotheses were generated initially from the 6WLC of SARS-CoV-2 NSP15 (crystal structure) and redocked structure using the default settings of Biovia DS 2020, respectively. The RLPG (Kurczab & Bojarski, 2013) method generates the model that corresponds to interactions of receptor-ligand complex. The complexes were aligned to each other and interaction frequencies of each residue with ligand were captured. Excluded volumes were generated automatically to prevent clashes. A set of pharmacophore models with maximum selectivity predicted by Genetic function approximation (GFA) (Rogers & Hopfinger, 1994) was selected.

Virtual screening & filtering of the hits
Virtual screening was carried out using the chosen pharmacophore model on a total of 19161 compounds of DrugBank, Shelleckchem database (Antiviral, Nucleotide Analogue, and Bioactive compounds) using the screen library module (Pal et al., 2019) of Biovia DS 2020. The hit compounds were sorted based on fit values and only those hits were considered for further investigation, which pass the pharmacophore screening (Fit value !3). To further refine the compounds, the Lipinski rule of five and the Veber rule were applied in Biovia DS 2020. Only those compounds were selected that passed the following criteria: molecule weight (MW) lesser than 500 g/mol, Hydrogen bond donors (HBD) lesser than or equal to 5, hydrogen bond acceptors (HBA) lesser than or equal to 10, LogP lesser than or equal to 5 and rotatable bonds lesser than or equal to 10 (Bickerton et al., 2012).
For predicting the pharmacokinetic properties of the hit compounds, ADMET (absorption, distribution, metabolism, excretion, and toxicity) profile was calculated by using the ADMET tool of Biovia DS 2020. This step is very crucial for predicting the future of compounds in the human body after drug administration. Next, the toxicology of the compounds was determined which is useful for lead optimization in Biovia DS 2020 by using the TOPKAT module based on Komputer Assisted Technology (Alam & Khan, 2018

Molecular docking-based virtual screening
A total of 97 compounds after the above filtration were subjected to virtual screening using molecular docking against the tertiary structure of SARS-CoV-2 Nsp15 (PDB ID: 6WLC). Molecular docking was performed by using CDOCKER (molecular dynamics simulated-annealing-based algorithm) program (Ding et al., 2020) in Biovia DS 2020. The binding site was defined based on co-crystallized ligand 'UMP' and grid sphere of 94.0389, À20.0126, and À25.899 points with the radius of 5.4852 Å was set. The following residues were defined within the binding site sphere: Gln245, Gly248, His250, Lys290, Cys293, Ser294, Pro341, Tyr343, Lys345, and Leu346, and rest of the docking parameters were set to be predefined (Kim et al., 2021). The best docking conformation for each hit compound was selected based on the lowest CDOCKER energy and interactions with key binding site residues. For cross-validation of molecular docking studies, we scored each ligand conformation by using the following conventional scoring algorithms: PLP1, PLP2 (CHEN et al., 2012), LigScore1, LigScore2 (Krammer et al., 2005), PMF, PMF04, Jain, Ludi Energy Estimate-1, Ludi Energy Estimate-2, and Ludi Energy Estimate-3 (Li et al., 2018) in Biovia DS 2020.

Consensus scoring
Consensus scoring merges more than two docking outcomes to give improved results of virtual screening analysis by increasing the probability of true positives and decreasing the possibility of false negatives. The selected 796 conformations of the candidate hits including the control ligand (UMP) from the molecular docking results followed by the ligand scoring method (second scoring) were used as input for the consensus scoring module in Biovia DS 2020. The molecules were sorted on the basis top rank percentile (Clark et al., 2002;Wang et al., 2003).

Molecular dynamics (MD) simulation
A 100 ns molecular dynamics (MD) simulation was done using Gromacs 2020.1 package with GROMOS96 forcefields to learn more about the dynamic nature and stability of protein-ligand complexes (Pronk et al., 2013). The top three ligands after docking into the Nsp15 binding site of SARS-COV-2 in addition to the control ligand were selected as a starting structure for MD simulations. All the complexes were defined in the cubic box with at least 1.0 nm distance from the defined complex system. Ligand's topologies were defined using Prodrg server (Sch€ uttelkopf & van Aalten, 2004) and were assembled into the protein topology file. Next, complexes were solvated using a simple point charge (spc) (Gl€ attli et al., 2003) water model followed by addition of sodium and chloride ions into the system for neutralization. Complexes were energy minimized using the steepest descent algorithm (50000 steps) to avoid any steric clashes.
To constrain the bond length, a LINear Constraint Solver (LINCS) function was adapted into the complete system and electrostatic calculations were performed using the particle mesh Ewald method (Shahbaaz et al., 2019). After the successful energy minimization step, NVT (constant number, volume, temperature) followed by NPT (constant number, volume, temperature) was done to equilibrate the system for 100 ps for the satisfaction of 1 atm bar pressure and 300 K constant temperature. Finally, 100 nanoseconds (ns) MD run were performed with an integration step of 2 fs. The stability, interactions, and fluctuation of the complex were studied by analysing the trajectory file of the system. Potential energy, root mean square deviation (RMSD), root mean square fluctuations (RMSF), the radius of gyration (RoG), and H-bond graph were plotted by using Grace Software (Abdul Samad et al., 2016).

Molecular mechanics Poisson-Boltzmann surface area (MM-PBSA)
To revalidate the MD simulation study, we enumerated the binding free energy of the top 3 selected hits along with the control compound 'UMP' in the presence of the Nsp15 of SARS-CoV-2 using 'Molecular Mechanics Poisson-Boltzmann Surface area' (MM-PBSA) of g_mmpbsa tool. We have extracted the MM-PBSA calculations from the last stable 20 ns of each MD run and adopted all the default parameters. All the calculations were done using the following equations as reported earlier (Kollman et al., 2000).

Results
Covid-19 is considered a global health pandemic. Presently, there is neither an effective treatment nor vaccine with full efficacy available to prevent SARS-CoV-2 viral disease. Nsp15 of SARS-CoV2 has emerged as a therapeutic target due to its important role in the replication of the virus. The crystal structure of this non-structural protein is present in the complex with UMP present in PDB (ID: 6WLC) (Kim et al., 2021), which opens the strategy to design inhibitors against Nsp15.
In this context, this study aims to utilize the structure-based pharmacophore model to virtually screen the new inhibitors which may target the Nsp15 with greater efficacy. Further MD simulation was performed to study the stability of the selected hit compounds complexed with Nsp15 of SARS-CoV-2. This perspective has been satisfyingly used to discover repurposed drugs/antiviral compounds in a cost and timeeffective manner.

Structure-based pharmacophore modeling
Pharmacophore illustrates the 3-dimensional positioning of important electronic and steric characteristics for the perfect binding of a hit compound to the protein. The pharmacophore model could be either ligand-based or structure-based depending upon the availability of known ligand activity or on available protein target information, respectively. In the current study, the receptor-ligand pharmacophore (structurebased) was built based on interactions between Nsp15 of SARS-CoV-2 and UMP. Before pharmacophore building, we calculated the ligand RMSD between co-crystallized ligand and redocked ligand (UMP) (Figure 2A), which was found to be 0.6979 A which proposes the docking accuracy of the software. It also suggests, the lowest energy conformation of UMP resembles the experimental mode of binding determined by crystallographic experiment. Overall, 9 features matched between receptor protein & the ligand, and 10 pharmacophore models were generated (Table 1). Out of 10 pharmacophore models, we selected Pharmacophore model-01 with the highest selectivity score of 12.137 and found it to possess six features set viz., AAADDD (three hydrogen bond acceptor and three hydrogen bond donor groups) ( Figure 2B and C). Excluded volumes were also added during pharmacophore generation which helps to protect the steric clashes between protein and ligand functional groups.

Virtual screening & filtering of the hits
The best pharmacophore model-01 was searched against a total of 19161 compounds of DrugBank, Shelleckchem database (Antiviral, Nucleotide Analogue, and Bioactive compounds) for the HTVS process. At the end of this procedure, we obtained 3523 hit compounds which were able to map on the pharmacophoric feature (AAADDD) with fit values varying between (4.98 to 2.56e-14 ). Out of 3523 compounds, 990 compounds having a fit value of more than 3.0 cut-off were selected after comparison with the control ligand 'UMP'. We then employed filters such as ADMET and Lipinski's-Veber for further filtering of our hits. These rules define the molecular properties of compounds that influence their pharmacokinetics and pharmacodynamics and affect their distribution, absorption, metabolism, toxicity, and excretion in the human body. Finally, 97 drugs were filtered with the help of ADMET, Toxicity (TOPKAT), and Lipinski Veber Filter ( Figure 1F).

Molecular docking & consensus scoring
We further investigated the important interactions against the key amino acid of the active site of the SARS-CoV-2 Nsp15 enzyme by docking through the CDOCKER tool of Biovia DS 2020. Out of 97 filtered compounds, 96 hit compounds successfully docked into the Nsp15 binding site of SARS-CoV-2. These 96 compounds were then sorted based on the CDOCKER Energy score. To remove any false positives in the dataset, we further calculated various empirical scores (PLP1, PLP2, LigScore1, LigScore2, PMF, PMF04, Jain, Ludi Energy Estimate-1, Ludi Energy Estimate-2, and Ludi Energy Estimate-3) for all the conformations of the ligands. Next, we calculated the consensus score which merges all the scoring functions as well as the CDOCKER energy score and helps in the significant enhancement of the hit rates. The docking, as well as consensus score of the top 10 hits, were enumerated in Table 2 & ST 2. All the ligand's hit conformations were ranked based on the consensus score, and we selected the top 3 hits with greater consensus score as compared to the control ligand (UMP). The top three candidates with high consensus scores toward the Nsp15 enzyme were found to be S5009/DB03312 (Consensus score: 12, Brivudine (BVDU), DB02980 (Consensus score: 12, Thymidine-5 0 -(dithio) phosphate) and DB03804 (Consensus score: 12, 5-Bromothienyldeoxy uridine) and their ADMET profile is enumerated in Table 3. Based on recent studies and co-crystallized ligand UMP (Kim et al., 2021), the active site of the protein consists of these amino acids: Gln245, His250, Lys290 Val292, Cys293,  Ser294, Thr341, Tyr343, Pro344, Lys345, and Leu346. As depicted in Figure Figure 3D). The interactions of these top hit compounds are also enumerated in Table  ST 3.

Molecular dynamics simulations
To further strengthen the virtual screening and molecular docking results, 100 ns MD simulation was performed with three hit compounds complexed with Nsp15 protein of SARS-CoV-2 along with the control ligand (UMP). MD simulation helps in the study of stability, energetics, and dynamic behaviour of complex protein with resulting trajectories in an environmental condition which is much like the in vivo conditions with respect to solvent, ions, pressure, and temperature. Throughout 100 ns MD run, receptor-ligand complexes were examined based on their average RMSD, ligand RMSD, RMSF, hydrogen bond analysis between protein as well as protein with ligand and radius of gyration (Rg).
The protein-backbone RMSD of the simulated system was retrieved from the output trajectory which explains the structural stability in the dynamic environment. Greater RMSD values represent the unfolding of the protein whereas lower values indicate compactness and equilibration in the protein structure. The RMSD graph for all the complexes was plotted and is given in Figure 4A. RMSD profile for Nsp15-screening hits (S5009, DB02980, and DB03804) was compared with the control ligand-complex (Nsp15-UMP) (Kim et al., 2021) as a reference. The RMSD plot between the backbone atoms of all the complexes was below 0.4 nm in the case of lead molecules, whereas the average RMSD in the case of the Nsp15control complex was greater than 0.6 nm suggesting that all the hit compounds were more stable as compared to the control-UMP. To gain more insights, we also calculated ligand rmsd and the results depicted that S5009 and DB03804 were more stable in terms of structural deviation whereas DB02980 was less stable as compared to the control ligand UMP ( Figure 4B).
Plots of RMSF were generated to evaluate the fluctuation of each amino acid residue which is responsible for dynamic system stability. The fluctuations were found to be almost similar for all the receptor-ligand complexes after studying the RMSF plot of each of the amino acid residues. There was not much fluctuation observed in the active site region (250-350) (Kim et al., 2021) of the protein (Nsp15) bound with either S5009, DB02980, and DB03804. However, the Nsp15-UMP complex showed fluctuations in the region of (198-209), (224-230), (312-316), and (335-338). Overall, RMSF values for lead molecules with Nsp15 were found to be more stable than Nsp15-UMP complex that represented the stability of these complexes ( Figure 4C). We next analyzed RoG using MD trajectories to study the steadiness in protein folding during MD simulation. The RoG value of all the complexes except the Nsp15-UMP complex was found to be very low and within the acceptable range without unusual or abnormal deviation ( Figure 4E). The hydrogen bond between protein and ligand is responsible for maintaining stable conformations throughout the MD simulation. Therefore, the H-bond plots over 100 ns of the MD simulation time for all the complexes were examined and are shown in (Figure 4E,F) which represented stability and rigidness in the proteinprotein and protein-ligand complex. The Nsp15-S5009 made the highest number of H-bonds (seven) in comparison to other lead molecules. Post-MD simulation, we extracted and analysed the final snapshot of the complexes and the analysis revealed that most of the active site residues remained intact into the binding site in all the complexes (Fig. S1).

Molecular mechanics Poisson-Boltzmann surface area (MM-PBSA)
The binding free energy for each complex was computed from the last 20 ns of the MD simulation trajectory. The g_mmpbsa tool was used from GROMACS implicit MM/PBSA technique to compute the binding energy. As depicted in Figure 5, the binding energy is negatively affected by electrostatic energy, van der Waal energy, and SASA energy, whereas the binding energy is positively affected by polar salvation energy. The binding free energy of each compound except DB02980 is lower than that of the control ligand, indicating that S5009 and DB3804 have better binding affinities against Nsp15-SARS-Cov2. However, Brivudine (S5009) is found to have the best binding energy (-71.029 kJ/mol) among all the top hits and interacted with Nsp15 with better affinity.

Discussion
The SARS-COV-2 endoribonuclease Nsp15 is an attractive therapeutic target for drug discovery against Covid19. However, unlike other drug targets like 3CLpro and RdRp, which are investigated extensively, and a large number of antivirals have been reported; Nsp15, has been comparatively a less explored drug target against the novel Covid-19. The traditional approach for antiviral drug discovery is a challenging task because it costs billions of dollars and usually takes many years of research. In silico drug, design provides another approach for rapid discovery of potential lead compounds and control of evolving diseases. In the present study, the pharmacophore model based on Nsp15-UMP interactions showed a good selectivity score, and virtual screening resulted in several molecules for repurposing. For further validation after docking studies on screened hits, we performed multi-score analysis and calculated the consensus score. The top 10 molecules were selected based on a consensus score. However, three (S5009, DB02980, & DB03804) out of 10 showed a similar comparative consensus score as shown by the control compound 'UMP'. Analysis of receptorligand interactions between Nsp15 and the top 3 hit compounds including the control compound, revealed that oxalone and pyrimidine rings were common among all the inhibitors, whereas the third hit compound (DB3804) had an additional thiophene ring which may be required for inhibitory activity of the compound. In all the three hit-compounds, most of the interactions were retained like control compound (UMP) and these hits were also found to be interacting with His235, His250, and Ser294 which are crucial for  Nsp15 activity of SARS-CoV based on previous mutational studies (Kim et al., 2021). Brivudine (S5009/DB03312) which comes as a common hit from DrugBank Database, nucleotide analog database, and antiviral database is a drug with antiviral activity and inhibiting viral replication by blocking the action of DNA polymerase (De Clercq, 2004). Thymidine-5 0 -(dithio)phosphate (DB02980) hydrolyses the DNA of the Penicillium citrinum which is an opportunistic pathogen and reported to cause pneumonia or skin-related disorders (Hesse et al., 2017;Overington et al., 2006). 5-Bromothienyldeoxyuridine (DB03804), the third top hit in our list of potential Nsp15 inhibitors, has previously been reported as a Human herpesvirus 1 (HHV-1) inhibitor and was also considered for Covid-19 (Imming et al., 2006). To cross-validate our pharmacophore-based virtual screening and molecular docking experiments, we conducted MD simulation and MM-PBSA calculations which led to the validation of three hits (S5009, DB02980, & DB03804) and that confirmed complex stability as well as good binding affinity with Nsp15. We also captured a snapshot ( Figure S2) from the MD trajectory of the top three hit compounds at 0, 25, 50, 75 and 100 ns to study the binding pose but no drastic changes were observed for the control compound (UMP), S5009, and DB03804. However, DB02980 shifted from the initial binding position which also correlated with the RMSD-ligand graph ( Figure 4B), Post-MD receptor-ligand interactions (Figure S1 C), and MM-PBSA profile ( Figure 5). These results could assist in understanding Nsp15 inhibition and will certainly channelize the gap between clinical results and preclinical test outcomes, and therefore, contribute significantly to the quick development of drugs to combat the novel coronavirus (SARS-CoV-2).

Conclusion
In the current study, we identified three drug molecules that can be repurposed as novel inhibitors for Nsp15 of SARS-CoV-2 as potential anti-SARS-CoV-2 hits by using the synergetic computational approach of structure-based pharmacophore modeling, virtual screening, molecular docking, and dynamics analyses. To make the study quick and effective, we retrieved antiviral, nucleotide analog, and repurposed compounds from the different databases and used graded filtering to shortlist compounds that can interact at the active site of Nsp15 with great affinity. We chose the top three ligands based on their consensus score. Next, the selected three ligands (S5009, DB02980, and DB03804) were further validated by molecular dynamics studies. Based on the RMSD analysis, H-bond analysis, Post MD interactions, and MM-PBSA calculations, Brivudine (S5009) an existing CMV (cytomegalovirus) inhibitor drug may be suggested a topmost lead for the Nsp15 of SARS-CoV-2. This study gave a preliminary basis that the hits obtained have the potential to inhibit NSP15 of SARS-CoV-2 based on in-silico studies. The screened lead molecules need to be further evaluated by using in vitro and in vivo studies to confirm their therapeutic status against COVID-19. The findings of this study may be relevant in the development of novel inhibitors based on Brivudine like scaffolds as an antiviral agent in the future.

Disclosure statement
There are no conflicts of interest declared by the authors.

Data availability statement
The datasets supporting the results and conclusion of this manuscript are included within the article and its supplementary information files.