Seawater fungi-derived compound screening to identify novel small molecules against dengue virus NS5 methyltransferase and NS2B/NS3 protease

Dengue fever is a virus spread by mosquitoes that has no effective treatment or vaccination. Several dengue cases combined with the current COVID-19 pandemic, exacerbates this problem. Two proteins, NS5 methyltransferase and NS2B/NS3 primary protease complexes, are crucial for dengue viral replication and are the target sites for antiviral development. Thus, this study screened published literature and identified 162 marine fungus-derived compounds with active bioavailability. Following Lipinski's rules and antiviral property prediction, 41 compounds were selected for docking with NS5 methyltransferase and NS2B/NS3 protease (PDB ID: 6IZZ and 2FOM) to evaluate compounds that could stop the action of dengue viral protein complexes. To find the best candidates, computational ADME, toxicity, and drug target prediction were performed to estimate the potential of the multi-targeting fungal-derived natural compounds. Analyzing the result from 41 compounds, Chevalone E (−13.5 kcal/mol), Sterolic acid (−10.3 kcal/mol) showed higher binding energy against dengue NS2B/NS3 protease; meanwhile, Chevalone E (−12.0 kcal/mol), Brevione K (−7.4 kcal/mol), had greater binding affinity against NS5 methyltransferase. Consequently, this study suggests that Chevalone E is an effective inhibitor of NS5 methyltransferase and NS2B/NS3 protease. Ligand-based virtual screening from DrugBank was utilized to predict biologically active small compounds against dengue virus NS2B/NS3 major protease and NS5 methyltransferase. Both licensed medications, estramustine (DB01196) and quinestrol (DB04575), were found to be similar to Chevalone E, with prediction scores of 0.818 and 0.856, respectively. In addition, cholic acid (DB02659), acitretin (DB00459), and mupirocin (DB00410) are similar to Sterolic acid, zidovudine (DB00495), imipenem (DB01598), and nadolol (DB01203) are similar to Brocazine A, and budesonide (DB01222) and colchicine (DB01394) are related to Brevione K. These findings suggest that these could be feasible dengue virus treatment options, meaning that more research is needed.

In recent years, marine fungi have been proven to be plentiful and a potential source of novel bioactive natural chemicals. Because most of these organisms exist in harsh environments, they can develop unique secondary metabolites. Metabolites are thought to represent the chemical defense response of fungi competing for substrates [12]. The range of natural chemicals produced by marine fungi suggests that some of these substances could be used in clinical trials to develop anti-infective medications.
Our study focused on using fungal bioactive compounds with potent antiviral activity against various known pathogenic viruses, such as influenza, hepatitis C virus, and herpes simplex virus, and as potential drug prototypes against DENV as a contribution to this global scientific endeavor. Thus, in this study, we used a systematic screening method to identify the top candidate for NS2B/NS3 and NS5 inhibitors of pathogenic DENV. Furthermore, computerized absorption, distribution, metabolism, excretion, and toxicity (ADMET) were analyzed to predict the potential of these multi-targeting fungal natural compounds for lead optimization and drug discovery.

Protein dataset and binding site analysis
X-ray crystallographic structures of nonstructural (NS) methyltransferase NS5 (PDB ID: 6IZZ) and NS2B/NS3 protease (PDB ID: 2FOM) were retrieved from the Protein Data Bank (PDB). Afterwards, active drug sites of two retrieved proteins were predicted using the Computed Atlas of Surface Topography of Proteins (CASTp) server for highresolution crystal structures and binding pockets [13].

Protein and ligand preparation
Before docking investigations, each protein was opened into the    Discovery Studio (DS) Visualizer, and water molecules, original inhibitor, and ligand compounds were removed to upgrade the protein's structure. The required hydrogen atoms were then inserted, followed by optimization to ensure structural stability. The final compounds were then translated to PDBQT format using the AutoDockTools-1.5.6 software.
To prepare the ligand, we examined a wide range of natural bioactive chemicals derived from marine fungi that have previously been outlined in the literature [14,15]. Subsequently, 162 compounds were enrolled with ID numbers and chemical structures obtained from the PubChem database (Supplementary Table 1). We then refined the bioactive compounds based on molecular weights between 350 and 500 (g/mol). Next, the biological potential of the selected compounds was predicted using the prediction service PassOnline server [16], which proposes the biological activity spectra of the compounds using the SMILES files of the structures. The likelihood of becoming active (pa) parameter was set to pa>0.3 for a better prediction against DENV. Finally, 41 substances were found to have antiviral properties. Then, each ligand's SMILE file was converted to PDBQT format, loaded into the AutoDockTools-1.5.6 software, and set up using the prepared ligand preparation tool for docking analysis.

Active site prediction and molecular docking
For molecular docking experiments, we used the default procedure in AutoDock tools 1.5.6 [17]. A grid box was created around the active site of DENV virus NS5 methyltransferase and NS2B/NS3 protease, as well as 2FOM and 6IZZ, with the help of a DS Visualizer. In addition, we set the grid box at 40 × 40 × 40 points in the xyz-dimension, which equaled a grid box spacing of 0.3753, and fixed the coordinates of the x, y, and z centers as 0.387667, − 6.969833, and 12.141667 for NS2B/NS3; meanwhile, − 15.658143, − 19.544214, and 38.773643 for NS5. We employed the Lamarckian genetic algorithm with default parameters for docking simulations, which included 10 genetic algorithm runs. Using the DS Visualizer, we further investigated the molecule with the highest energy ranking for protein-ligand interactions.

Drug profile analysis of top compounds
Absorption, distribution, metabolism, and excretion (ADME) are the four key criteria that determine drug levels and kinetics of drug exposure in an organism's tissues. These characteristics play a major role in the pharmacological activity and performance [18]. The SwissADME server was used to evaluate the ADME properties of the top five metabolites [19]. The blood-brain barrier (BBB) in the examined substances was calculated using the BOILED-Egg model [20].

Toxicity, carcinogenicity, and mutagenicity prediction
To predict toxicity, mutagenicity, and carcinogenic effects, canonical SMILES of the selected compounds that displayed predicted antiviral activity were uploaded to the pkCSM server [21] and proTox-ll databases [22]. The toxicity was predicted using the toxicity mode in the pkCSM server, and the ProTox-ll server was used to assess carcinogenicity and mutagenicity. By combining the molecular similarity, fragment tendency, and fragment similarity approaches, this popular server effectively predicts numerous toxicity outcomes [23]. Based on the analysis of 2-dimensional (2D) similarity to substances with a known median lethal dose, the server also projected the oral toxicity (LD50). The list used for the prediction contained almost 38,000 different chemicals with known oral LD50 values in mice [24].

Prediction of drug target and available drug molecules from DrugBank
The SwissTargetPrediction server was used to determine the potential macromolecular targets of the therapeutic candidates [25]. Based on a combination of 2D and 3D similarities with a library of 370,000 known bioactive chemicals on approximately 3000 proteins, the server makes predictions. Analyzing the homology screening of anticipated top drug candidates, the SwissSimilarity web tools were utilized to find possible therapeutic compounds against DENV virus NS5 methyltransferase and NS2B/NS3 protease. Using diverse methodologies, such as FP2 fingerprints, electroshapes, and spectrophores [26], the server allows ligand-based virtual screening of several libraries of small compounds to locate authorized, investigational, or commercially accessible medications from DrugBank.

Molecular dynamics simulations
The iMOD server was used to analyze the structural dynamics of the best protein-ligand combination and to evaluate the stability of the structure by applying deformability analysis, which included the computation of the eigenvalues of the complexes [27,28]. This is a web server that can be customized and can generate complex deformability, variance, B-factor, covariance map data, and elastic network data. The deformability of a complex or protein is determined by its ability to deform at each amino acid residue [29]. The eigenvalue is equal to the energy difference required to bend a particular structure; therefore, the smaller the eigenvalue, the easier it is to deform the complex. The eigenvalue also represents the stiffness of the protein complex [30].

Analysis of drug surface hotspot and ligand binding pocket
The drug surface hotspots of the selected NS5 methyltransferase and NS2B/NS3 protease with the ligand were investigated based on the structural conformation of the docked complexes. The ligand-binding patterns as well as the locations of the interacting residues are shown in Table 2. The positions of amino acids 127-161 were found to be crucial for NS2B/NS3 binding interactions (2FOM). Contrarily, the amino acids from positions 806 to 840 showed the interactions with NS5 (6IZZ) protease. Additionally, CASTp was used to confirm the binding site residues of the two target proteins.

The binding affinities of the ligands into NS2B/NS3 and NS5 active site
The results of docking details were recovered after docking all the ligands with NS2B/NS3 protease and NS5 methyltransferase target in Table 1. The molecules that had the lowest binding energy of the docking score were considered the best molecules and had a higher binding affinity with the target receptors [31]. Moreover, the lower the stabilization energy of ligand binding to the receptor, the greater the potential of action was. To limit the probability of false-positive results, the ideal binding energy of marine fungal-derived compounds was compared with that of the approved clinical chemical treatments, and the binding energy in the screening criterion was amended to a negative value > − 5 kcal/mol [32] to reduce the risk of false-positive results. The four best components are presented in Table 2, and the docking scores of all the compounds are listed in Table 1 In the previous study, the in-silico highest binding affinities of Artesunic acid and Homoegonol against the NS5 methyl transferase reported as − 7.2 and − 7.1 kcal/mol [33], which refers the slightly higher binding affinity energy of our selected compounds (Chevalone E, Brevione K, Brocazine A, and Brevione F). In another study were revealed that apigenin and luteolin phytocompounds showed the highest binding affinity (− 7.7 kcal/mol) against dengue NS2B/NS3 protease [34] that was dissimilar from our results we found more binding affinity − 13.5, -10.5, − 9.2 and − 9.0 kcal/mol for Chevalone E, Sterolic acid, Brevione K, Brevione F, respectively. In addition, [35] reported a lower binding affinity for phytocompounds based on the binding affinity of the above-listed compounds. The ligand-receptor interactions are shown in Figs. 1 and 2.

Hydrogen bond analysis
The complex interactions between NS2B/NS3 and NS5 with inhibitory ligands were visualized using a DS Visualizer, and a hydrogen bond analysis was also performed. In the Table 2, the numbers of hydrogen bonds and residues implicated in hydrogen bond interactions were summarized. Chevalone E, Sterolic acid, Brocazine K, Brevione F, and lindgomycin are possible NS2B/NS3 inhibitor candidates with drug-like properties and no toxicity, carcinogenicity, or mutagenicity. At the residues of SER127, LEU128, GLY153, and HIS51, Sterolic acid forms four hydrogen bonds, whereas only the ARG54 residue was formed in the case of Brocazine K. Moreover, Brevione F and Lindgomycin (GLY151: 2.37) showed that they could also be used as NS2B/NS3 inhibitors. Contrarily, at the residues TYR838 and ILE717, Brocazine A formed hydrogen bonds with viral NS5 methyltransferase. In addition, Sterolic acid had three hydrogen bonds interacting with the active site of NS5 methyltransferase at residues ALA757, TRP803, and GLU509.
Interestingly, despite the fact that Brevione K had a higher binding affinity, this combination was unable to form hydrogen bonds. Therefore, Brocazine A, with two hydrogen bonds, was found to be the best inhibitor of NS5 methyltransferase. The common interactions of the examined compounds, when compared to the original inhibitor, suggest that marine-derived fungal compounds Chevalone E, Sterolic acid, Brevione K, and Brocazine A (Fig. 3) could be potential inhibitors of NS2B/NS3 protease and NS5 methyltransferase.

In silico evaluation of drug likeness and ADME
The ADME approach was used to analyze the drug-likeness of the top four compounds found in the marine-derived fungi. Forecasts were made using the SwissADME database. The likeness of the drug was demonstrated by five rules established by Lipinski. The molecular weight (MW) should be between 350 and 500 (g/mol), the number of hydrogen bond acceptors should be between (10), the number of hydrogen bond donors should be between (5), and the Log Po/w should be between 5 and 10 [36], with no more than one violation allowed. Chevalone E, Brevione K, and Sterolic acid had higher gastrointestinal absorption than Brocazine A. Furthermore, the BOILED-Egg model was used to compute BBB penetration, which demonstrated that none of the top medication candidates tested had BBB permeation. Each compound was water soluble to varying degrees, with Sterolic acid having the    highest solubility (Table 3). Therefore, these molecules can be used to mimic the effects of drugs (Fig. 4).

Toxicity, carcinogenicity and mutagenicity prediction
Acute toxicity, hepatotoxicity, carcinogenicity, mutagenicity, immunotoxicity, and toxicity targets were explored as the toxicity endpoints. The results revealed that Chevalone E and Brevione K fell in the category of toxicity class 4, while Sterolic acid and Brocazine A showed toxicity levels of 1 and 3, respectively (the lower the class, the higher the toxicity). The estimated LD50 for Chevalone E, Sterolic acid, Brevione K, and Brocazine A, were 1600, 34, 1255, and 75 mg/kg, respectively. The toxicity characteristics in (Table 4) depict the level of confidence in the positive toxicity results compared to the class average. No unfavorable effects such as tumorigenicity, mutagenicity, irritation, or reproductive consequences were observed for any of the compounds.

Prediction of drug targets, available drug molecules from DrugBank and molecular dynamics simulation
Molecular target investigations are required to uncover the phenotypical side effects or possible cross-reactivity caused by their actions. This is the first step in determining the viability of using the drug in future in vitro and in vivo experiments [37]. As a result, Fig. 5 shows the top 25 findings generated by SwissTargetPrediction [38] for Chevalone E (NS2B/NS3 and NS5 inhibitor), Sterolic acid (NS2B/NS3 inhibitor), and Brevione K (NS2B/NS3 and NS5 inhibitor). The target sites that the chemical could bind to were largely protease and enzyme (13.3%) for Sterolic acid, nuclear receptor (33.3%) for Chevalone E, and enzyme (26%) for Brevione K (Table 5).
Ligand-based virtual screening was used to predict biologically active small molecules against NS2B/NS3 protease and NS5 methyltransferase of dengue from DrugBank. Estramustine (DB01196) and Quinestrol (DB04575), both licensed medications, were similar to Chevalone E, with prediction scores of 0.818 and 0.856, respectively. Furthermore, the results revealed that cholic acid (DB02659), acitretin (DB00459), and mupirocin (DB00410) are similar to Sterolic acid; zidovudine (DB00495), imipenem (DB01598), and nadolol (DB01203) are similar to Brocazine A; and Budesonide (DB01222) and Colchicine (DB01394) are similar to Brevione K (Table 6). These findings indicate that these could be viable therapeutic candidates for DENV infection, implying that further research is needed. In addition, based on molecular dynamics modeling, the best complexes for NS5 methyltransferase-Chevalone E and NS2B/NS3 protease-Chevalone E showed satisfactory eigenvalues with structural stability. The maximal eigenvalue of the NS2B/NS3 protease-Chevalone E complex was 2.866006e-04, compared to 7.701529e-05 for the NS5 methyltransferase-Chevalone E complex, showing great flexibility and difficulty in deformation ( Fig. 6 and Fig. 7).

Conclusions
Currently, there are no specific treatments for dengue. Although few studies have been conducted to develop dengue vaccines, this process has a great deal to be accomplished. As a result, scientists are currently looking for inhibitors to block DENV's key methyltransferase and protease, NS5 and NS2B/NS3, which are involved in viral replication, as targets for next-generation therapeutic development. On the contrary, marine fungi have proven to be a rich and promising source of novel bioactive natural compounds and may be able to suppress viral NS5 and NS2B/NS3 activities. In this study, the primary viral protease was docked against marine-derived chemicals. In addition, drug-like characteristics, toxicity, carcinogenicity, and mutagenicity were predicted using in silico ADME analysis. Analyzing the results, our study suggests that Chevalone E, Sterolic acid, Brevione K, and Brocazine A found in marine-derived fungi are the best NS5 and NS2B/NS3 inhibitors. The predicted drug based on the ligand estramustine, cholic acid, acitretin, colchicine, and zidovudine could be exploited and developed as an alternative or complementary therapy for the treatment of dengue virus.

Funding information
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.