In Silico Exploration of Potential Natural Inhibitors against SARS-Cov-2 nsp10

In continuation of our previous effort, different in silico selection methods were applied to 310 naturally isolated metabolites that exhibited antiviral potentialities before. The applied selection methods aimed to pick the most relevant inhibitor of SARS-CoV-2 nsp10. At first, a structural similarity study against the co-crystallized ligand, S-Adenosyl Methionine (SAM), of SARS-CoV-2 nonstructural protein (nsp10) (PDB ID: 6W4H) was carried out. The similarity analysis culled 30 candidates. Secondly, a fingerprint study against SAM preferred compounds 44, 48, 85, 102, 105, 182, 220, 221, 282, 284, 285, 301, and 302. The docking studies picked 48, 182, 220, 221, and 284. While the ADMET analysis expected the likeness of the five candidates to be drugs, the toxicity study preferred compounds 48 and 182. Finally, a density-functional theory (DFT) study suggested vidarabine (182) to be the most relevant SARS-Cov-2 nsp10 inhibitor.


Introduction
More than 217 million humans around the world were confirmed to be infected with COVID-19 and another 4.5 million families lost one of their beloveds as stated by the WHO on 2 September 2021 [1]. In response, all scientists in the field of drug discovery should work unceasingly to discover a cure against the notorious virus.
Computer-assisted (based or aided) drug design is a well-established branch of drug design that covers various in silico computational and theoretical approaches. These approaches are essential contributors to the development of new bioactive agents [2][3][4][5][6][7][8].
SARS-CoV-2 is an enveloped positive-sensed RNA virus. The replication of SARS-CoV-2 depends on a group of 16 non-structural proteins. These proteins have the codes of nsp1-nsp16. Between them, the two proteins nsp10 and nsp16 make an essential protein complex [51]. That complex is responsible for the vital methylation reaction at the ribose 2′-O position of the penultimate nucleotide of the viral RNA cap [52]. Accordingly, if a molecule could bind with that enzyme and inhibit this essential step, the replication process will be stopped.
The targeting of SARS-CoV-2 nsp-16 with a library of 10 [53] and 265 [54] FDA-approved compounds was studied before. Likely, a group set of 22 natural compounds from some Indian plants was computationally screened against six non-structural-proteins of SARS-CoV-2 [55].
In this study, different computational (in silico) selection methods were applied to 310 candidates. The examined candidates were chosen through a deep database search according to three parameters. The first parameter was to be naturally isolated. The second was having exhibited antiviral potentiality before. Lastly, we considered that the culled compounds belong to different chemical classes and accordingly have various chemical structures. The applied computational techniques were a structural similarity study against SAM followed by a fingerprint study against the same target. The selected candidates were docked against nsp10 (PDB ID: 6W4H) to prefer 44, 48, 85, 102, 105, 182, 220, 221, 282, 284, 285, 301, and 302. Then ADMET and toxicity studies further picked two candidates. Finally, a DFT study suggested the most relevant inhibitor of SARS-Cov-2 nsp10 (Figure 1).

Molecular Similarity against SAM
The basic principle of 2D Molecular similarity is that molecules with similar chemical structures are expected to have similar biological activities [56].
To measure the similarity of two objects, their general features have to be compared. On a molecular level, the molecular features or descriptors of any compound start from

Molecular Similarity against SAM
The basic principle of 2D Molecular similarity is that molecules with similar chemical structures are expected to have similar biological activities [56].
To measure the similarity of two objects, their general features have to be compared. On a molecular level, the molecular features or descriptors of any compound start from the general physicochemical properties and extend to more specific structural features such as partition coefficient (ALog p) [57], molecular weight (M. Wt) [58], hydrogen bond donors (HBA) [59], hydrogen bond acceptors (HBD) [60], number of rotatable bonds [61], number of rings, and also aromatic rings [62], in addition to molecular fractional polar surface area (MFPSA) [63].
All mentioned molecular properties were used in the applied similarity study between the natural candidate's set ( Figure S1, Supplementary Materials) and the co-crystallized ligand (SAM) of SARS-CoV-2 nonstructural protein (nsp10) (PDB ID: 6W4H) using Discovery studio software. Thirty candidates ( Figure 2) were chosen to be the most similar to SAM. number of rings, and also aromatic rings [62], in addition to molecular fractional polar surface area (MFPSA) [63].
All mentioned molecular properties were used in the applied similarity study between the natural candidate's set ( Figure S1, Supplementary Materials) and the co-crystallized ligand (SAM) of SARS-CoV-2 nonstructural protein (nsp10) (PDB ID: 6W4H) using Discovery studio software. Thirty candidates ( Figure 2) were chosen to be the most similar to SAM.
As shown in Figure 2, the similar candidates showed a high degree of structural similarity with SAM. In detail, most candidates have a sugar-like moiety as that of SAM as candidates 85, 102, 105, 120, 182, 183, 203, 204, 220, 221, 282, 284, 285, 301, and 302. These moieties may serve as a good center for hydrogen bonding interaction with the target receptor. Furthermore, most candidates have hetero bicyclic structures as present in SAM. Besides, xanthine-like structures were defined in many similar candidates such as 182, 284, 285, and 301.  As shown in Figure 2, the similar candidates showed a high degree of structural similarity with SAM. In detail, most candidates have a sugar-like moiety as that of SAM as candidates 85, 102, 105, 120, 182, 183, 203, 204, 220, 221, 282, 284, 285, 301, and 302. These moieties may serve as a good center for hydrogen bonding interaction with the target receptor. Furthermore, most candidates have hetero bicyclic structures as present in SAM. Besides, xanthine-like structures were defined in many similar candidates such as 182, 284, 285, and 301.
As shown in Figure 3, the candidate's set was divided into six smaller sets. From the first set to the fifth comprised 50 candidates while the sixth set was 60. As shown in Figure 3, the candidate's set was divided into six smaller sets. From the first set to the fifth comprised 50 candidates while the sixth set was 60.

Filter Using Fingerprints
The fingerprint is another similarity technique that depends on the 2D molecular structures of two different ligands in a binary format. This technique computes the presence and/or absence of several sub-structural fragments to calculate the degree of intermolecular structural similarity. This technique is utilized as a tool to detect the degree of similarity between a hit candidate and a lead one [64] The fingerprint approach examines the following parameters: charges [65], hybridization [66], H-bond acceptors, and donors [67], positive and negative ionizable moieties [68], halogens, and aromatic rings beside the ALogP category of candidates. The experiment was carried out using Discovery Studio.
The fingerprint's output depends on Tanimoto coefficient (SA/(SA + SB + SC)). SA is a symbol that represents the number of bits present in the reference molecule (SAM) and the examined candidate. On the other hand, SB and SC represent the number of bits in the examined candidate but not SAM and the number of bits in SAM but not the examined candidate, respectively. The Tanimoto coefficient gives values with a range of zero (no shared bits) to one (all bits the same).
The reported antiviral potentialities of the preferred metabolites were summarized in the Supplementary Materials.
The docking process was validated through a redocking step of SAM against active pockets of SARS-CoV-2 nonstructural protein (nsp10). The suitability of the performed protocol was demonstrated by the small RMSD (0.60 Å) that was found between the docked pose and SAM ( Figure 4). The reported antiviral potentialities of the preferred metabolites were summarized in the Supplementary Materials.
The docking process was validated through a redocking step of SAM against active pockets of SARS-CoV-2 nonstructural protein (nsp10). The suitability of the performed protocol was demonstrated by the small RMSD (0.60 Å) that was found between the docked pose and SAM ( Figure 4).  The mode of binding of SAM inside COVID-19 nsp10 was illustrated in Figure 5. It was noticed that SAM interacted with the active site via the formation of six hydrogen bonds with Lys6844, Leu6898, Asn6899, Asp6912, Cys6913, and Tyr6930.
The methylpyrimidine-2,4-dione derivative (220) possessed a good potential affinity of −21.17 into the COVID-19 nsp10 active site. This high affinity is attributed to the formation of five hydrogen bond interactions. The pyrimidine moiety of candidate 220 was involved in two hydrogen-bonding interactions with Asp6912 and Cys6913. While the furan part interacted with the active site by three hydrogen bonds with Leu6898 and Tyr6930 ( Figure 6). The mode of binding of SAM inside COVID-19 nsp10 was illustrated in Figure 5. It was noticed that SAM interacted with the active site via the formation of six hydrogen bonds with Lys6844, Leu6898, Asn6899, Asp6912, Cys6913, and Tyr6930. Among all studied metabolites, members 220, 48, 182, 221, and 284 exhibited the greatest binding free energies of docking (Table 3). The methylpyrimidine-2,4-dione derivative (220) possessed a good potential affinity of −21.17 into the COVID-19 nsp10 active site. This high affinity is attributed to the formation of five hydrogen bond interactions. The pyrimidine moiety of candidate 220 was involved in two hydrogen-bonding interactions with Asp6912 and Cys6913. While the furan part interacted with the active site by three hydrogen bonds with Leu6898 and Tyr6930 ( Figure 6).   Candidate (48) exhibited a binding mode like that of SAM with the formation of four hydrogen bonds with Cys6913, Tyr6930, and Leu6898 (Figure 7).   Investigation of the top docking poses of the 6-aminopurine member (182) showed that it interacted with the COVID-19 nsp10 active site through the formation of three hydrogen bond interactions. Its amino group was involved in a hydrogen bond with Asp6912 while one purine nitrogen atom formed a hydrogen bond with Cys6913. In addition, the furan oxygen interacted by a hydrogen bond with Tyr6930 ( Figure 8). Investigation of the top docking poses of the 6-aminopurine member (182) showed that it interacted with the COVID-19 nsp10 active site through the formation of three hydrogen bond interactions. Its amino group was involved in a hydrogen bond with Asp6912 while one purine nitrogen atom formed a hydrogen bond with Cys6913. In addition, the furan oxygen interacted by a hydrogen bond with Tyr6930 ( Figure 8).
The proposed binding pattern of the pyrimidinedione derivative (221) was illustrated in Figure 9. It interacted with the active site via the formation of five hydrogen bonds with Asn6899, Asp6897, Cys6913, and Tyr6930. The proposed binding pattern of the pyrimidinedione derivative (221) was illustrated in Figure 9. It interacted with the active site via the formation of five hydrogen bonds with Asn6899, Asp6897, Cys6913, and Tyr6930.  The proposed binding pattern of the pyrimidinedione derivative (221) was illustrated in Figure 9. It interacted with the active site via the formation of five hydrogen bonds with Asn6899, Asp6897, Cys6913, and Tyr6930.

In Silico ADMET Analysis
Five parameters were examined for candidates 48, 182, 220, 221, and 284 using Discovery studio software. Acyclovir, the potent anti-viral drug, was used as a reference candidate. The results are illustrated in Figure 11.
All the tested candidates have a very low chance to penetrate BBB. This indicates the high safety margin of such derivatives against the CNS. Additionally, all candidates exhibited an aqueous solubility character. For intestinal absorption, candidates 48, 182, 220, and 221 were predicted to have poor to very poor levels, while candidate 284 was expected to have a moderate level. Furthermore, all candidates were predicted to be CYP2D6 noninhibitors and can bind plasma protein by less than 90%. These results indicated that all the tested candidates have good pharmacokinetic properties and can be utilized for further investigations.

In Silico ADMET Analysis
Five parameters were examined for candidates 48, 182, 220, 221, and 284 using Discovery studio software. Acyclovir, the potent anti-viral drug, was used as a reference candidate. The results are illustrated in Figure 11. high safety margin of such derivatives against the CNS. Additionally, all candidates exhibited an aqueous solubility character. For intestinal absorption, candidates 48, 182, 220, and 221 were predicted to have poor to very poor levels, while candidate 284 was expected to have a moderate level. Furthermore, all candidates were predicted to be CYP2D6 noninhibitors and can bind plasma protein by less than 90%. These results indicated that all the tested candidates have good pharmacokinetic properties and can be utilized for further investigations. Figure 11. The expected ADMET study. Figure 11. The expected ADMET study.
All the tested candidates have a very low chance to penetrate BBB. This indicates the high safety margin of such derivatives against the CNS. Additionally, all candidates exhibited an aqueous solubility character. For intestinal absorption, candidates 48, 182, 220, and 221 were predicted to have poor to very poor levels, while candidate 284 was expected to have a moderate level. Furthermore, all candidates were predicted to be CYP2D6 noninhibitors and can bind plasma protein by less than 90%. These results indicated that all the tested candidates have good pharmacokinetic properties and can be utilized for further investigations.

In Silico Toxicity Studies
Candidates 48, 182, 220, 221, and 284 were tested in silico for their proposed toxicity using Discovery studio software. In this test, seven toxicity models were utilized using ribavirin as a reference. The results are summarized in Table 4. FDA rodent carcinogenicity in female mice indicated that candidates 48 and 182 were non-carcinogenic, while candidates 220, 221, and 284 had some sort of carcinogenicity. Besides, candidates 48, 182, and 284 showed TD 50 values of 9.295, 4.245, and 6.402 mg/kg body weight/day, respectively. Candidates 220 and 221 showed high carcinogenic potency TD 50 values of 67.851 and 55.437 mg/kg body weight/day, respectively. Furthermore, candidates 48 and 182 showed high rat maximum tolerated dose values of 0.191 and 0.175 g/kg body weight, respectively. On the other hand, candidates 220 and 221 showed low rate maximum tolerated dose values of 0.095 and 0.094 g/kg body weight, respectively. Candidate 284 showed a comparable rat maximum tolerated dose value (0.155 g/kg body weight) with ribavirin (0.154 g/kg body weight). The tested candidates showed rat oral LD 50 values ranging from 0.778 to 6.173 g/kg body weight, which were higher than the reference drug LD 50 = 0.750 g/kg body weight. For the rat chronic LOAEL model, candidates 48 and 182 showed high values of 0.018 and 0.010 g/kg body weight, while candidates 220, 221, and 284 showed low values of 0.009, 0.006, and 0.004 g/kg body weight, respectively. All candidates were predicted to have mild to moderate irritant effects against ocular irritancy and skin irritancy models. Accordingly, candidates 48 and 182 had low toxicity profiles and were preferred for further studies.

DFT Studies
DFT parameters (Table 5) were studied for candidates 48 and 182 [69,70] against SAM as a reference using Discovery studio software ( Table 5, Figures 12 and 13). Orbital (MO) analysis of EHOMO represents the energy of the highest occupied molecular orbital. On the other side, ELUMO represents the lowest unoccupied molecular orbital energies. The MO analysis is one of the essential parameters that is linked to the chemical reactivity and stability of a molecule. The HOMO spatial distributions of SAM are mainly presented on the 2-aminobutanoic acid moiety (the electron transfer zones), while its LUMO spatial distributions are located on the tetrahydrofuran-3,4-diol moiety (the electron acceptor zones). For candidate 48, the HOMO spatial distributions are mainly located on the (2R,3R,4R)-2-(hydroxymethyl)pyrrolidine-3,4-diol moiety, while its LUMO spatial distributions are found on the (S)-pyrrolidin-3-ol moiety. For candidate 182, the HOMO spatial distributions are mainly presented on the 9H-purin-6-amine moiety, while its LUMO spatial distributions are located on the (2R,3S,4R)-2-(hydroxymethyl)tetrahydrofuran-3,4diol moiety. Furthermore, the gap energy of candidate 182 (0.128 kcal/mol) was less than that of candidate 48 (0.210 kcal/mol), confirming the high reactivity of candidate 182. Consequently, candidate 182 may serve as a promising candidate for further studies.

Molecular Electrostatic Potential Maps (MEP)
MEP was used to specify the electrostatic potential of 48, 182, and SAM in a 3D form via the calculation of the partial charges, electronegativity, and chemical reactivity [71]. The electrostatic potential affects the binding of a drug with a specific protein and gives a deeper insight into drug-receptor interaction [72]. In MEP, the red color denotes the electronegative atoms, which can go through hydrogen bonding interactions as an acceptor. Additionally, the blue color denotes the electron-poor atoms that can form a donor in hydrogen bonding. The green to yellow color denotes the neutral atoms, which can form hydrophobic interactions [73].

Name
Total  The MEPs of SAM, 48, and 182, were illustrated in Figure 13A, B, and C, respectively. Investigating these figures indicated that SAM has eight red patches that are suitable for hydrogen bonding acceptors and are considered favorable sites for the electrophilic attack. Also, it comprises six blue patches that are suitable for hydrogen bond donors (the most favorable sites for the nucleophilic attack). Candidate 182 has six red patches and five blue patches. In addition, there is a yellow patch on the 9H-purine nucleus indicating a high possibility for hydrophobic interaction. These findings are highly like that of SAM. The MEP of candidate 48 is slightly different from SAM. In detail, it has four red patches and four blue patches. These results indicated that candidate 182 has a greater similarity with SAM than candidate 48. Because of that, candidate 182 was singled out.
tack. Also, it comprises six blue patches that are suitable for hydrogen bond donors (the most favorable sites for the nucleophilic attack). Candidate 182 has six red patches and five blue patches. In addition, there is a yellow patch on the 9H-purine nucleus indicating a high possibility for hydrophobic interaction. These findings are highly like that of SAM. The MEP of candidate 48 is slightly different from SAM. In detail, it has four red patches and four blue patches. These results indicated that candidate 182 has a greater similarity with SAM than candidate 48. Because of that, candidate 182 was singled out. The antiviral activities of the preferred candidate, vidarabine (182), were reported against several viruses in different reports. It was active against herpes simplex encephalitis and neonatal herpes simplex infection [74,75], HBV [76], varicella-zoster virus [77], human polyomavirus [78], adenovirus [79], and Epstein-Barr virus infection [80].

Molecular Similarity Detection
Achieved by Discovery studio software (see method part in Supplementary Materials).

Pharmacophoric Study
Achieved by Discovery studio software (see method part in Supplementary Materials).

Docking Studies
Docking studies were achieved by MOE.14 software (see method part in Supplementary Materials).

ADMET Analysis
Achieved by Discovery studio 4.0 (see method part in Supplementary Materials).

DFT Studies
Achieved by Discovery studio software [84] (see method part in Supplementary Materials).

Conclusions
Vidarabine (182) was suggested to be the most relevant SARS-Cov-2 nsp10 inhibitor among 310 naturally isolated metabolites that exhibited antiviral potentialities before. This suggestion was based on different computational (in silico) selection methods that included molecular similarity assessment, molecular fingerprint, docking studies, toxicity, ADMET, and DFT. The selected candidate showed various antiviral activities before. Further in vitro and in vivo biological studies have to be conducted to confirm the effect of 182 against SARS-Cov-2 nsp10 and its potential as an anti-COVID-19 drug.
Supplementary Materials: The following are available online, Figure S1: Chemical structures of the examined natural antiviral compounds, Table S1: Detailed toxicity report, in addition to the method (Molecular Similarity, Pharmacophore, Docking studies, ADMET studies, Toxicity studies and DFT studies).