Virtual screening and molecular docking of structures with potential inhibitor of the ebolavirus glycoprotein

Ebola is a disease caused by viruses of the ebolavirus genus that cause hemorrhagic fever with high human mortality. Symptoms of the infection include fever, headache, joint pain, diarrhea, vomiting and stomach pain. Unfortunately, there is still no effective treatment for this disease. This work aimed to perform a virtual screening of candidates with EBOV GP inhibition potential based on pharmacophores of 4 structures from the literature. The selected candidates underwent ADME analysis and later calculations of HOMO, LUMO and electrostatic potential map. 50 structures were obtained and after selection based on their ADME characteristics only 4 candidates were chosen. In the molecular anchoring process, the structure that presented the best binding energy was ZINC12 with -4.44 kcal.mol Research, Society and Development, v. 11, n. 2, e45311226034, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i2.26034 2 and enzyme inhibition constant Ki equal to 555.9 μM. This result is fundamental, as it will subsidize other researches such as in vitro studies, or even application in real samples.


Introduction
Ebola is a disease caused by viruses of the ebolavirus genus that cause headache, joint and muscle pain, diarrhea, vomiting, stomach pain. In some cases, those infected may present with a rash and bleeding, in addition to hemorrhagic fever with high human mortality (Khabbaz et al., 2015;Breman et al., 2016). These viruses were discovered in the 1970s and are part of the Filoviridae family, infectious viruses such as Marburg and Cueva-virus. This family is known for its filamentous characteristics (Figure 1), are negative strand RNA viruses (Rougeron et al., 2015;Towner et al., 2007), are highly pathogenic and are classified as "risk class 4" of according to the World Health Organization (WHO) (World Health Organization, 2020; Mali & Chaudhari, 2019;Baize et al., 2014;Feldmann & Geisbertet, 2011).  Society andDevelopment, v. 11, n. 2, e45311226034, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i2.26034 3 Recent Ebola cases were presented by the WHO demonstrating that between 2018 -2020 in the Democratic Republic of Gong Zaire there were 3,481 new cases, causing the death of 2,299 people, thus presenting 66% fatality during this period (World Health Organization, 2020). During these outbreaks the effectiveness of some drugs were tested to fight the virus, such as clinical trials in Guinea. The rVSV-ZEBOV vaccine was studied as a way to prevent new outbreaks of the disease, people who received the vaccine did not show infection during 10 days of investigation, in the same period of time 23 people who were not vaccinated showed infection. This vaccine is still being used to control the disease in the Republic of Congo, therefore, the WHO still recognizes the need to evaluate new candidate vaccines to combat the Ebola virus (World Health Organization, 2020). While there are several candidates for an Ebola vaccine (Richman & Nathanson, 2016), there is still no evidence of effective treatment for people already afflicted with the disease.
It is believed that the Ebola virus can be transmitted between humans and/or animals through contact with the body fluids of those already infected. An illustrative scheme of the possible viral cycle of the disease is illustrated in Figure 2.
Humans and other mammalian species are definitive hosts of the Ebola virus, and endemic areas are investigative foci of the disease. (World Health Organization, 2020;Feldmann & Geisbertet, 2011). In view of studies carried out to identify the natural reservoirs in Ebola hemorrhagic fever outbreaks, experiments carried out with African plants and animals have shown that rodents and bats are possible foci of the Ebola virus (Baize et al., 2014). Source: Massachusetts Medical Society, available on the site: https://biology.stackexchange.com/ Four drugs, benzatropine, bepridil, paroxetine and sertraline were investigated by Ren et al. (2018) for the inhibition of Ebola virus infections with action on glycoprotein. These drugs already exist on the market, and therefore, the side effects, pharmacokinetics and pharmacology are already known.
Benzotropine is an anticholinergic drug, known as an acetylcholine receptor antagonist (Jacoby, 2017), is a drug used for the treatment of symptoms related to Parkinson's disease (Johns, 2014), and cocaine addicts, as it acts as a dopamine inhibitor based on GBR12909 (Rothman et al., 2008).
The drug bepridil is used in the treatment of dysrhythmia by blocking the sodium current and expelling calcium from cardiac cells. One of the adverse effects of the use of this drug is a ventricular arrhythmia ("spins of the tips") (Desai, 2016). In Research, Society andDevelopment, v. 11, n. 2, e45311226034, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i2.26034 4 an isolated case, pneumonia caused by the use of bepridil was observed in a 78-year-old man (Yokoi et al., 2013).
Paroxetine and Sertraline are inhibitors that act on serotonin reuptake acting as a drug with antidepressant action, their performance is low in terms of noradrenaline reuptake (Van Harten, 1993). Both drugs generate collective effects such as nausea, dizziness, insomnia and irritation (Desai, 2016). The use of paroxetine associated with tamixifen (estrogen receptor) causes a decrease in this, as it acts as a strong inhibitor of cytochrome P450 2D6, the study was carried out with 30 breast cancer survivors (Loprinzi et al., 2011).
Virtual screening is a computational methodology aimed at accelerating the discovery of new drugs. This computational methodology uses several methods to evaluate and direct choices of chemical structures with modulated biological applications, that is, it facilitates the scientific investigation of possible drug candidates (Dos Santos et al., 2020;Wermuth et al., 2015).
Molecular Docking uses computational processes that provide estimates of ligand-protein free energy. For this, several spatial conformations of the complex are obtained, making it possible to evaluate the most stable conformation (lower energy) and the greater inhibitory effect (higher enzyme inhibition constant) (Souza, 2015).
Computational chemistry has brought innovations to several fields of investigation, its contributions include several theoretical approaches, using methods ranging from mechanical calculations to ab initio calculations such as the Density Functional Theory. It can be used to obtain the molecular optimization of different structures, obtain vibrational spectra, evaluate bond angles and distances, thus aiming to explore all possible molecular characteristics through computational tools (Ault, 2018;Larkin, 2017;Nova & Maseras , 2013).
In this work, a pharmacophore-based virtual screening was performed to search for potential inhibitors of EBOV GP, responsible for binding to the host cell receptor that leads to membrane fusion (Ren et al., 2018). In addition, the pharmacokinetic and toxicological properties of the molecules selected in the virtual screening were investigated. For the best candidates, the interaction between the ligand and the target active site was investigated (PDB code 6F6S available in the Protein Data Bank) (Berman et al., 2002) via molecular anchoring in order to evaluate the binding energy (ΔG) , enzyme inhibition constant (Ki) and molecular interactions of ligands with the target protein.

Molecule selection and computational calculations
The structures of 4 drugs ( Table 1) that present studies related to inhibition of Ebolavirus infections (Ren;Zhao;Fry;Stuart, 2018) were selected through the DRUGBANK (Wishart et al., 2018) server to serve as a basis for pharmacophore modeling.
The structures were designed in 2D using the free software MarvinSketch 20.7 (Csizmadia, 1999) and optimized using the free program ArgusLab (Naz et al., 2009) using the semi-empirical Hamiltonian PM3 (Anders;Koch;Freunscht, 1993) method to obtain the best structural conformation. The HOMO, LUMO orbitals and the electrostatic potential map were calculated for the selected candidates (Braga et al., 2016). Source: Own Authorship.

Pharmacophore Modeling and Virtual Screening
After optimization, the files were aligned in Discovery Studio software version 2.1 (Studio, 2008) to obtain similar characteristics and atomic positions, using the crystallographic structure of Benzatropine as a pivot, and sent to the PharmaGist web server (Schneidman-Duhovny et al. al., 2008) to generate at least 3 pharmacophores, which was used to deduce the main structural features through molecular characterization to perform the biological activity (Arooj et al., 2013).
With the pharmacophoric resources generated, the search for coordinates and visualization took place using the Accelrys Discovery Studio software (Studio, 2008), through the Pharmit web server (Koes, 2018) the virtual sorting was performed based on the points through the coordinates, and the max total hits to get the 50 best fit structures.

ADMET Predictions
The properties of Absorption, Distribution, Metabolism and Excretion (ADME) together with the Toxicological ones were performed on the preADMET web server (Lee et al., 2003), through similarity calculations and pharmacokinetic profile of other drugs (Da Silva Costa et al., 2018;Cruz et al., 2018;Vieira et al., 2014). The drugs Bepridil and Paroxetine were used as benchmarks.

Molecular Docking
The target protein EBOV GP (PDB: 6F6S) (Ren;Zhao;Fry;Stuartet, 2018) was worked on at UCSF Chimera version 1.15 (Pettersen et al., 2004) removing water molecules and other residues that could interfere with the ligand-macromolecule interaction. To validate the active region of the PDB, molecular redocking of the benzatropin ligand complexed to the target protein was performed (Da Silva Costa et al., 2018).
The molecular anchorage calculation was carried out with the help of the free software AutoDock Tools (Morris et al., 2009), where hydrogen was added to the ligands and receptors, performing the partial calculation of gasteiger charges (Gasteiger & Marsili, 1980). The grid box of size 22x16x28 with spacing of 0.375 Å was defined based on the size of the ligand structures, being centered at the coordinates X=-47.084, Y=11.220 and Z=-3.788. Such coordinates were obtained through the redocking process close to the MET458 residue of the B chain. The simulation was carried out with the Lamarckian genetic algorithm (Morris et al., 1998), with the ligand considered flexible in relation to the rigid target protein.
The other values were assigned by default. The enzymatic complexes obtained from the anchoring process were organized into families according to the RMSD -Root Mean Square Deviation. The best conformations in terms of energies (Bastos et al. 2020) were chosen for the discussion of the results. Hydrophobic interactions and hydrogen bonding were analyzed with the software Discovery Studio and UCSF Chimera.

Molecular Dynamics
The four structures studied in molecular docking were submitted to the later stage of the study using the methodological approach of Rajapaksha et al., (2020) Yang et al., (2020) and Oso et al., (2020).
Conventional molecular dynamics (CMD) simulation was performed through the LARMD server (http://chemyang.ccnu.edu.cn/ccb/server/LARMD/index.php/home/index) (Yang et al., 2020) loading the protein-ligand complex from the molecular docking results. Standard parameters of 4 ns were used for MD and water time as an explicit model. The simulations were performed by the AMBER16 program using AMBER ff14SB (Maier et al., 2013;Plewczynski et al., 2011) and General AMBER Force Field (GAFF) (Wenthur et al., 2014;Wang & Merz, 2006) for the ligands and waste.
The pdb complexed with the ligands was solvated in a TIP3P water octahedron box (Wold et al., 2018;Price & Brooks, 2004), for the counter-ions system Na+ or Cl-were used.
The minimization energy occurred using the SANDER module of the AMBER16 software (Yang, 2015;Case et al., 2005) with 2000 steep descent steps and 3000 conjugate gradient steps, using the PMEMD module (Yang, 2015;Case et al., 2005). In the first stage, the system underwent heating from 10 to 300 K in 30 ns. Then, the release process similar to minimization took place, the atoms were relaxed at 300K and 1 atm. The Cpptraj module of the AMBER16 program (Roe & Cheatham, 2013;Case et al., 2005) was used for the analysis of RMSD, Rg and RMSF.

Synthetic Route in silico
The theoretical synthetic route was aided by the investigation of the retro synthesis using the Spaya server (Parrot et al., 2021), its platform results in real synthesis proposals classified through score, where score 1 represents a synthesis 100% similar to the literature reaction.

Pharmacophore Modeling and Virtual Screening
After molecular optimization, the pharmacophore model of drug structures was obtained through alignment and evaluation by the PharmaGist server, which presented a score of 4.815. This resulted in 3 pharmacophoric characteristics: an aromatic structure, a hydrogen acceptor and a hydrophobic one ( Figure 3). The possible regions where biological interactions can occur between the ligands and the target are described via coordinates according to table 2. Source: Own Authorship.
The virtual screening performed by the Pharmit server resulted in the best 50 structures that fit the pharmacophore region, so that after visual analysis, only 25 candidates remained to move on to the next phase of the study. Source: Own Authorship.

ADMET Prediction
The results of the in silico pharmacokinetic predictions were separated into Absorption (Table 3), Distribution (Table   4) and Metabolism (Table 5) Human intestinal absorption (HIA) is of great importance because it acts directly on drug transport in the body, and is also used to predict the oral availability of some drugs (Yan et al., 2008). Molecules that present values lower than 0.4 x 10-6 cm/s present low oral absorption. In relation to this parameter, we can classify molecules as malabsorption (0-20%), moderate absorption (20-70%) and good absorption (70-100%) (Castillo-Garit et al., 2008). All selected molecules obtained good results in human intestinal absorption (HIA) tests, presenting values above 90%, with a variation between 93.525% and 100% (Table   3).
Cell permeability Caco is directly related to the intestinal barrier used to predict drug absorption (Sambuy et al., 2005;Balimane;Chong;Morrison, 2000). Values below 4 nm/s indicate low permeability, between 4 and 70 nm/s medium permeability and greater than 70 nm/s high permeability in relation to PCaco (cell permeability) (Yamashita et al., 2000;Irvine et al., 1999). All molecules showed medium permeability. In relation to PCaco-2, Paroxetine had the highest permeability value of 57.13 nm/s, and the lowest value was for the ZINC1 molecule with 34.74 nm/s. The values for permeability in human skin (Pskin) were negative for all molecules, so the administration by application to the skin should not be significant. According to Irvine et al., (1999) the MDCK cell system can be used as a tool for rapid screening for permeability.
Values below 100 nm/s are considered to have low permeability (Dezani et al., 2016). All screened molecules showed high permeability ZINC1 (307.78 nm/s), ZINC4 (220.15 nm/s), ZINC12 (188.64 nm/s), ZINC20 (173.62 nm/s) and ZINC14 (139.12 nm/s), while the drugs Bepridil and paroxetine showed low permeability of the MDCK cellular system. Table 4 shows the distribution of properties in percentage of Plasma Protein Binding -PPB and P-glycoprotein, the main contributors to the evaluation of drug distribution properties. Binding to plasma proteins is one of the mechanisms for the drug to be distributed in the human body via the bloodstream, providing a better therapeutic effect (Moda, 2007). PPB values (>70%) show significant protein binding (Benet & Hoener, 2002 The blood-brain barrier (BBB) is related to the blood-brain interface that maintains the balance of the central nervous system (CNS), circumscribes the introduction of drug administration (Mendonça, 2016). Compounds with BBB values lower than -1 present poor distribution to the brain, whereas BBB values greater than 1 have an easy crossover and present Central Nervous System activity (Ma;Yang, 2005). For values related to BBB, the structure of the molecule ZINC14 (0.88) and ZINC1 (0.69) present the best values among the 5 candidates, approaching the reference drug Paroxetine (0.27). The molecules ZINC4 (2.18), ZINC12 (2.15), ZINC20 (1.18) showed values greater than 1, characterizing a high probability of penetration (Mendonça, 2016). The drug Bepridil obtained a high value of 5.99 in relation to penetration of the blood-brain barrier. Note: a) percentage of binding to plasma proteins; b) penetration of the blood-brain barrier (Blood-Brain Barrier-BBB); c) Inhibition of P-Glycoprotein. Source: Own Authorship.
The P-glycoprotein indicates a defense against chemical agents that are not usual in the body (Azeredo;Uchôa;Costa, 2009) increasing the resistance of cells (Reeve, 2014). No molecule showed inhibition of P-glycoprotein activity.
The CYP450 enzymes (CYPs) are related to xenobiotic compounds that support the induction of more than one CYP450 gene family. The liver is the organ where CYP450 plays an important role, as it is where it is active in the catalysis of biotransformation related to xenobiotics (Ortiz et al., 2012;Orellana & Guajardo, 2004). All screened molecules showed inhibition of cytochrome P450 2C19 enzymes, but drugs did not show inhibition, drugs and molecules inhibited cytochrome P450 2C9 (Table 5).  (Ortiz et al., 2012). All selected molecules showed inhibition of cytochrome P450 3A4 enzymes. Cytochrome P450 enzymes metabolize most drugs used in the abrogation of neoplasms or malignant cells (Marcolin et al., 2004).
The Ames test consists of using standardized cultures of Salmonella typhimurium, functioning as a drug receptor to cause distinct mutations. This test is used to analyze the mutagenicity of drugs, being one of the most prescribed by world entities (Santos, 2007). Table 6 presents the data related to mutagenicity, carcinogenic effects in rats and mice. All molecules showed mutagenicity in relation to the Ames test. In the specific tests through the TA 100 and TA 1535 strains analyzed, the effectiveness or non-efficacy of the molecules in relation to mutagenicity can be verified. The drugs Bepridil and Paroxetine showed more violations in total compared to the investigated molecules, thus demonstrating that even drugs on the market have aspects that violate the standards.

Computer Calculations
Molecular orbitals have important characteristics for biological systems, such as orbital density, charge and volume, as they directly influence the ability of the ligand to interact with the enzymatic active site (Pereira et al., 2016). The HOMO and LUMO boundary molecular orbitals of each of the structures can be seen in Figure 4. The HOMO orbital of the ZINC1 structure is balanced in almost all molecules, only the sulfur atom of the tetrahydrothiophene fragment and the oxygen of oxacyclopentane did not present orbitals. The LUMO for this molecule has a higher density in the region of the nitrogen double bond. ZINC4 features the HOMO and LUMO on the tetrahydrothiophene fragment on opposite sides of the structure.
For the ZINC12 molecule, the HOMO is concentrated towards the oxacyclopentane fragment with the exception of the   Source: Own Authorship.

Validation and Molecular Docking
The validation of the active site and the molecular anchoring process occurred through the evaluation of the Root Mean Square Deviation (RMSD) between the crystallographic benzantropine ligand with the pose obtained from the computational ligand. Figure 6 demonstrates the overlap of both ligands, which presented a result of 1.8232 Å, to be considered satisfactory, the RMSD must present a result of 2 Å or less (Ramírez & Caballero, 2018).  KJ.mol -1 . This type of energy occurs when a ligand-protein complex is formed, they are characteristics for analyzing the ease of protein-ligand interaction, the lower the energy value, the more likely a binding to occur (Souza, 2015;Gujjula, 2008). The selected structures, despite not having better binding energies than the comparison drugs, have favorable energy for an interaction to occur, as the binding energy score usually occurs between -7 to -10 KJ/mol (Azam & Abbasi, 2013;Bissantz et al., 2000).
The ZINC12 molecule showed the best binding energy with a value of -18.59 KJ.mol -1 and an inhibition constant equal to 555.9 µM. No hydrogen bonds were observed for this molecule, but it showed 12 hydrophobic interactions between the ligand and the target protein. The amino acid MET548 performed a pi-sulfur interaction, the stacked pi-pi interaction occurred with the amino acid TYR517, LEU515 performed a pi-sigma type interaction, the pi-alkyl interactions occurred with residues ALA101, LEU184, LEU43, LEU554, LEU558, VAL66, residues ILE38, ILE555 and LEU186 performed van der Waals interactions (Figure 8). Source: Own Authorship.
The ZINC20 molecule did not form a hydrogen bond with the EBOV GP receptor. The molecular interactions that occurred were of the pi-alkyl and alkyl type at amino acids ILE38, LEU184, LEU43, LEU515, LEU554, LEU558, MET548, TYR517, VAL66 (Figure 9). Affinity parameters for ZINC20 molecule were 989.31 µM for inhibition constant and binding energy of -17.17 KJ.mol -1 .  Research, Society andDevelopment, v. 11, n. 2, e45311226034, 2022 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v11i2.26034 15 The chemical interactions of the ZINC14 ligand as alvo will consist of a pi-enxofre interaction with the amino acid MET548 and pi-alkyl with the residues LEU554 and LEU558 (Figure 11). In addition to these, interactions with 10 hydrophobic residues occur and the formation of a hydrogen bond with a distance of 5.34 Å as the amino acid ILE185 in the oxadiazole region where it presents the LUMO orbital and the most electronegative area with no potential map. The binding energy obtained in the interaction of this ligand with the EBOV GP receptor was -16.37 KJ.mol -1 and an inhibition constant of 1350 µM. The ZINC4 molecule presented the highest inhibition constant equal to 1820 µM, however it was the structure with the lowest binding energy of -15.66 KJ.mol -1 , with 9 hydrophobic interactions and forming a hydrogen bond with the amino acid LEU554 with distance 4.40 Å. The TYR517 residue performed a stacked pi-pi interaction, MET548 presented a pi-sulfur interaction, the amino acids LEU184, LEU43, LEU515, LEU558 and VAL66 showed pi-alkyl interactions ( Figure 12).

Molecular Dynamics
To evaluate the conformational stability of the receptor-ligand complex, 4 ns molecular dynamics simulations were applied to the EBOV GP receptor and the ligands ZINC01, ZINC04, ZINC12, ZINC14 and ZINC20, observing the conformational changes and binding affinities. the main characteristics of ligands that affect conformation can be observed in the high fluctuations in root mean square deviation (RMSD) and root mean square fluctuation (RMSF) (Mukund et al., 2019).
In figure 13 we can see how the different ligands indicate variations in fluctuations, thus verifying that they have a critical role in the process of conformation movement. The structures of ZINC20 and ZINC04 have a lower RMSD value compared to the others, which can confirm that the structure of EBOV GP is more stable with these ligands. Source: Own Authorship.
The ZINC14 ligand showed little conformational fluctuation, obtaining its stability from 2500 ps, while the ZINC20 ligand presents some fluctuation peaks, its greater stability occurs between 1000 and 2500 ps, the others present large variations in their stabilities.
The by residue, Root Mean Square Fluctuations (RMSF) (Figure 14) showed fluctuation between 10 and 60 Å, we can observe that the ligand ZINC20 was the one that presented the lowest values of fluctuation, but still being close to the others as shown in the figure 14.  Note: Electrostatic energy (ELE), van der Waals contribution (VDW), total gas phase energy (GAS), non-polar and polar contributions to solvation (PBSOL / GBSOL), Final estimated binding free energy (deltaPB / deltaGB) calculated from the above terms (PBTOT / GBTOT) and entropy (TS), their unit is kcal / mol. Source: Own Authorship.
It is noticeable that the contributions of van der walls (VDW) and the total energies of the gas phase (GAS) were the ones that contributed the most to the interaction energy of the ligands with the receptor, whereas the entropy energies were the ones that contributed the least to the binding free energy result. ZINC20 showed the highest number of hydrogen bonds during molecular dynamics, interacting with 8 residues ALA101 ( The molecular dynamics results confirm the binding free energy and the tendency of these ligands to bind with the EBOV GP receptor, demonstrating that they are potential targets for future investigations.

Synthetic Route in silico
The ZINC1 showed a score of 0.8 in the retrosynthesis obtained from the SPAYA server, the theoretical synthesis of the compound ( Figure 15) starts with the production of forfural (I), it can be obtained from the dehydration of xylose (Mamman et al., 2008;Zeitsch, 2000), may also occur by the putative mechanism for the acid-catalyzed dehydration of xylose, in which displacement of the protonated C-2 hydroxyl group leads to a 2,5-anhydride intermediate which dehydrates to furfural (Nimlos et al., 2006;Antal Jr et al., 1991). Subsequently, it is necessary to obtain the compound 1,3-thiazolidin-2-ylhydrazine (II) from 1,3-thiazol-2-amine following the reaction observed in Singh et al., (2005). The ZINC1 molecule can be obtained in step III with the reaction of forfurral (I) with 1,3-thiazolidin-2-ylhydrazine (II) through the open vessel process and cooling during the heating of the synthesis (La Regina et al., 2011). Source: Own Authorship.
The retrosynthesis score of the ZINC4 compound was 1 and can be seen in figure 16, the first step of the synthetic route consists of obtaining 2-(Bromomethyl)furan (I) from the methodological procedure patented by Beaulieu & Ohemeng, (1999). In sequence, the synthesis of the compound ZINC4 (II) is carried out by mixing phenol with 2-(Bromomethyl)furan (I), a process already known in the literature (Annales de chimie 1942). Source: Own Authorship.
The synthetic route of the compound ZINC12 is similar to that of the compound ZINC1 (Figure 17), in this route only one synthesis step is necessary, the reaction of forfural with aniline according to the literature procedures (Matusiak et al., 2013), its score in the retro synthesis was 0.8.

Figure 17
-Theoretical route to obtain the ZINC12 compound.
The synthetic route proposed for the compound ZINC14 ( Figure 18) occurs through the reaction of the solution of (1,3,4-Oxadiazol-2-yl)methanol added to phosphorus oxychloride and the mixture is stirred at room temperature for a period of 6 hours (Pahutski et al., 2015) resulting in 2-(Chloromethyl)-1,3,4-oxadiazole which in turn used the methodology of Mohanazadeh e Aghvami (2007) to obtain the compound ZINC14. His score on the spay server was 0.8.

Figure 18
-Theoretical route to obtain the ZINC14 compound.
The score obtained for the compound ZINC20 was 0.9, its route is similar to the route of the compounds ZINC1 and ZINC12 where forfural is part of the reaction with 3-Hydroxypyridine to obtain the compound studied ( figure 19) based on the methodology of catalyzed oxidative esterification obtained from the literature (Zhang et al., 2011).

Figure 19
-Theoretical route to obtain the ZINC20 compound.

Conclusion
Through the virtual screening procedure based on pharmacophores, it was possible to obtain several structures where 5 of them (ZINC1, ZINC4, ZINC12, ZINC14 and ZINC20) were selected for further investigation.
ADMET's predictions showed that the structures are similar to the commercial drugs used in the comparison.
Molecular docking indicated the ZINC12 molecule as the most promising in this study, presenting a binding energy of -18.59 KJ.mol -1 , favoring its interaction with the Ebolavirus glycoprotein and an inhibitory constant of 555.9 µM. The molecular dynamics study thus demonstrating its inhibitory potential for EBOV GP. The synthetic route showed ways to obtain the compounds that could be tested.
Future researches are needed to prove the action of the selected structures, in vitro studies would be recommended considering the structures demonstrated here with virus cell cultures by specialized laboratories.