Molecular docking based design of Dengue NS5 methyltransferase inhibitors

Dengue is a viral infection caused by RNA infection of the family Flaviviridae and spread by the Aedes mosquitoes. Dengue NS5 methyltransferase is a known drug target for the disease. Therefore, it is of interest to design potential inhibitors for the target using molecular docking analysis. Our analysis shows the binding of compounds STOCK1N-98943, STOCK1N-98872, STOCK1N-98956, STOCK1N-98865, and STOCK1N-98950 with the protein drug target with optimal binding features for further in vitro and in vivo evaluations.


Background:
The dengue infection is caused by Flavivirus of the family Flaviviridae and it is an arthropod-borne diseases that consolidates four particular serotypes (DEN-1, DEN-2, DEN-3, and DEN-4) [1, 2]. The World Health Organization (WHO) considers dengue as a significant overall general prosperity challenge in the tropic and subtropics nations. Dengue is seen as a result of an unnatural climate change, unconstrained urbanization, inefficient mosquito control, and non appearance of social insurance facilities. [3][4][5]. Two and a half billion individuals live in dengue-endemic regions [6] and about 400 million contaminations occurring every year, with a death rate out-performing 5-20% in some areas [7]. Dengue disease is seen in excess of 100 nations including Europe and the United States (USA) [2]. The essential and non-fundamental proteins of DENV have been focal point of an antiviral structure. The DENV non-assistant proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5) have role in replication and virion assembly [8]. Screening of compounds as antiviral drugs is gaining momentum in recent years. Peptide-based drugs with higher bioactivity and low lethality has also been considered [9]. Compounds with anti-viral effects for the avian flu infection subtype H9N2 [10] and subtype H5N1 [11] are also known. Therefore, it is of interest in this context to screen for compounds against the Dengue NS5 methyltransferase as potential inhibitors for consideration as anti-viral drugs.

Three-dimensional structure prediction by I-TASSER:
The sequence of 262 aminoacid residues of methyltransferase domain (MTD) was retrieved from the Swiss Prot database in FASTA format. The three-dimensional model was generated using the I-TASSER server by multiple threading alignments and iterative structural assembly simulation [15].

Validation of the predicted model:
The Ramachandaran plot further validated the conformation of the best model predicted by the I-TASSER. The model quality was calculated by analyzing the phi (Φ) and psi (Ψ) torsion angles using the PROCHECK server. The Ramachandaran plot obtained from PROCHECK help select a good quality model with over 90% residues in the favoured region [16].

Prediction of the active site of the methyltransferase domain:
The active sites in MTD were identified using the computed atlas of surface topography of proteins (CASTp) server (http://cast.engr.uic.edu). It defines all the possible pockets in the protein structure. It measures the area and volume of each pocket and cavity analytically, both in solvent accessible surface and molecular surface [17][18][19]. Here, we input the target protein for predicting the ligand binding sites, and the CASTp server predicts the amino acids crucial for binding interactions and docking studies. 24 Ident1 is the percentage sequence identity of the templates in the threading aligned region with the query sequence. Ident2 is the percentage sequence identity of the whole template chains with query sequence. Cov represents the coverage of the threading alignment and is equal to the number of aligned residues divided by the length of query protein. Z-score is the normalized Z-score of the threading alignments. Alignment with a Normalized Z-score >1 mean a good alignment and vice versa. 977 TM-score is a recently proposed scale for measuring the structural similarity between two structures. RMSD a is the RMSD between residues that are structurally aligned by TM-align. IDEN a is the percentage sequence identity in the structurally aligned region. Cov represents the coverage of the alignment by TM-align and is equal to the number of structurally aligned residues divided by length of the query protein. 396 ©Biomedical Informatics (2019)

Selection of ligands and protein target molecule:
The schematic diagram of the workflow design is demonstrated in Figure 1. Ligand molecules were selected from the IBS natural compound library (InterBioScreen Ltd). Ligands were prepared using the LigPrep module of the Maestro 10.5 application. LigPrep performs many corrections on the ligands. These include the addition of hydrogens, 2D-3D conversion, corrected bond lengths and bond angles, low-energy structure and ring conformation. Allatom force field charges and atom types were assigned by the optimized potential for fluid simulations (OPLS_2005) force field [20-23]. One conformation for each ligand was thus generated for docking. The structural model of the NS5 methyltransferase generated by ITSER was used in this study. The Maestro 10.5 protein preparation module was used where changes such as the addition of hydrogen atoms, assigning bond orders, creation of zero-order bonds to metal, creation of disulfide bonds, fixing of the charges and orientation of groups were incorporated into the structure.

Molecular docking:
Molecular docking studies using the selected ligand molecules were completed using Maestro 10.5 molecular docking suite [13, 18, 24-26]. Each of these compounds was docked into target protein accordingly with positions, orientations, and conformations of the ligand in the receptor binding site, and the docking structure is possessing the lowest energy was preferred. In the present study, we screened approximately 50,000 natural compounds from the IBS against NS5 methyltransferase. IBS natural compounds docked with every selected protein molecules by using HTVS. To provide a better correlation between right poses and functional scores, GLIDE-XP mode was subsequently used on the selected conformations in HTVS mode. From Gscore, we select 20 compounds for GLIDEXP molecular docking. After the completion of ligands and proteins preparation, a receptor grid file was generated. For running the grid generation module, we have scaled van der Waal (vdW) radii of receptor atoms by 1.00 Ǻ as the default setting of Maestro 10.5. The active site of the receptor maintain a precise scoring function with thermodynamically most favourable energy and is calculated on a grid by various sets of fields. After the formation of the receptor grid file, flexible ligands with rigidreceptor-based molecular docking were performed. The best-fit compounds have been chosen for each target by optimal energy value and types of interactions.
Absorption, distribution, metabolism, excretion, and toxicity (ADME/T) properties ponders: Most of the medication applicants don't prevail in clinical preliminaries because of poor toxicology evaluations (ADME/T). In this way, ADME/T properties of best-docked mixes were selected utilizing QikProp utilization of Maestro 10.5 (auxiliary, physicochemical, biochemical, pharmacokinetics, and lethality properties). It predicts inborn properties of the atoms (medicate like properties, for example, octanol/water segment, log BB, by an overall CNS activity, register IC50 for Herg K+ channel blockage, Caco-2, MDCK cell porous features and logKhsa for human serum albumin binding [27, 28].

Results and discussion: Sequence analysis and domain association:
We have analyzed the amino acid sequence of the methyltransferase domain (262 aminoacid) by using different available tools. Figure 1A shows the Dengue virus (DEV) NS5 methyltransferase domain architecture. The InterProScan results show that methyltransferase (MTs) constitute an important class of enzymes present in every life form ( Figure 1B). They transfer a methyl group most frequently from S-adenosyl L-methionine (SAM or AdoMet) to a nucleophilic acceptor such as nitrogen, oxygen, sulfur or carbon leading to S-adenosyl-L-homocysteine (AdoHcy) and a methylated molecule. The crucial amino acids for binding are shown using different colours ( Figure 1C).

Three dimensional structure prediction:
The predicted model of MTD protein and its three-dimensional coordinate file in PDB format were successfully obtained from I-TASSER. The results obtained from the server includes predicted secondary structure with a confidence score (range 0 to 9), predicted solvent accessibility, five predicted structures with Cscore, top ten templates from PDB used in alignment, high ten PDB structural analogs, functional analogs protein, and binding site residues. Model MTD (Figure 1 D) was selected as the bestpredicted model with C-score 0.99, TM-score 0.85±0.08, and RMSD 2.8±2.1Å. C-score with higher value reflects a model of better quality. Top ten threading templates for query protein sequence MTD were identified by LOMETS meta-server (Table 1). Normalized Z-score generally estimates the threading alignment. However, a normalized Z-score >1 value reflects a confident, but in case of small alignment of the large query sequence, it does not give a significant indication of modelling accuracy. The percentage sequence identity in the threading aligned region (Iden1) and in the whole chain (Iden2) considered for the excellent homology. The structural alignment program, TM-align, identified 2px5A in PDB 397 ©Biomedical Informatics (2019) library as best structural analog of the top scoring model of I-TASSER with the TM-score of 0.792,0.979and 0.979respectively (Table 2).

Assessment of predicted model:
The Ramachandaran plots of the best-predicted model were obtained from PROCHECK servers which showed the reliability of the model. The PROCHECK Ramachandaran plot showed 98.2% residues in most favored regions and 0.5% residues in additional allowed regions, i.e., the total of 98.7% residues in allowed regions which indicates a good quality model of MTD (Figure 1 E).

Screening of targeted compounds against MTD protein:
In the 21st century has the disadvantage that it is challenging to get new compounds into the clinic. The target compounds were identified after virtual screening (VS). This is a reliable, inexpensive method for identifying leads by MTD. Figure 2 clearly reveals that complete flow chart of compound screening.

Dynamic site:
The protuberant binding site of MTD was calculated using the CASTp server with ideal parameters (Figure 3). CASTp evaluation observed the active site amino acids, surface area (473.597) and volume (558.187) of MTD. In MTD protein, all 62 binding pockets were categorized to find the residues with probe 1.4Å radius. The Cyan color denotes the active site amino acid residues involved in the binding pockets ( Figure 3A & B).

Molecular docking analysis:
This study is to identify the potent inhibitor against NS5 methyltransferase using molecular docking. Molecular docking of NS5 methyltransferase against natural compounds has been completed. First, we performed HTVS of IBS against the NS5 methyltransferase shown (Table 6). Further, 20 best compounds that possess minimal G score were performed using the XP mode of GLIDE. Our result shows that compounds with good dock score for protein NS5 methyltransferase (  STOCK1N-98950 showed the best G score -8.24, -7.41, -6.95, -6.73 and -6.52 kcal/mol, respectively. The molecular docking study illustrates the protein-ligands interactions and to summarize the various bonds such as hydrogen and electrostatic bond. STOCK1N-98943 was found to be the most potent and nicely bounded into the active site of NS5 methyltransferase with best G score compared to Chloroquine (Table 3)   Glide extra precision scores (kcal/mol); Lipophilic E Vdw; Chemscore lipophilic pair term and fraction of the total protein-ligand vdw energy; HBond; Hydrogenbonding term; Electro; Electrostatic rewards; Protein ligands interaction; p-p stacking, p-cat interaction and hydrogen bond between the ligands and protein .967 Predicted IC50 value for blockage of HERG K+channels; (acceptable range above -5.0)Molecule STOCK, InterBioScreen's library (IBS), Q P log Poct; was predicted partition coefficient of octanol/gas,(8.0 to 35.0); QPP Caco, predicted apparent Caco-2 cell permeability in nm/s. Caco-2 cells is a model for the gut blood barrier (nm/s)<25-poor,>500great. Q P log BB, predicted brain/blood partition coefficient; QPP MDCK, predicted apparent MDCK cell permeability in nm/s. MDCKcells are considered to be a good mimic for the blood-brain barrier; (nm/s)<25-poor,>500-great; Q P log KP, Predicted skin permeability; QP log Khsa Prediction of binding to human serum albumin; (acceptable range -1.5 to 1.5)

ADME properties:
Pharmacokinetic and pharmacodynamics properties of lead compounds were evaluated using the Qikprop application of Maestro 10.5. Compounds STOCK1N-98943, STOCK1N-

Biological activity predictions:
The selected bioactive constituents were evaluated for possible biological activity using the PASS online server. The biological activity spectrum (BAS) of a compound is known to have pharmacological effects, specific toxicities, and mechanisms of action. Because these probabilities can be calculated independently, the Pa and Pi values vary from 0 to 1, and Pa + Pi < 1. Pa is for the class of active compounds and Pi is for stable compounds [29]. PASS prediction results showed that Pa value is higher than Pi value inferring anti-dengue activity of selected compounds (Table  5). It should be noted that all compounds have shown a significant Pa value compared to Pi value with the potential for inhibiting NS5 methyltransferase.

Conclusion:
Dengue has progressed as a common disease affecting about 2.5 billion individuals over 100 countries. Therefore, it is of interest to design potential inhibitor drugs to control and combat Dengue fever. We analyze known structural data using molecular docking and report the binding properties of potential inhibitors to dengue NS5 methyltransferase for further consideration. These inhibitors have optimal binding features with the drug target.

Conflict of interest:
The authors declare no conflicts of interest.