The Crimean-Congo Hemorrhagic Fever virus (CCHFV) is a segmented negative single-stranded RNA virus (−ssRNA) which causes severe hemorrhagic fever in humans with a mortality rate of ~50%. To date, no vaccine has been approved. Treatment is limited to supportive care with few investigational drugs in practice. Previous studies have identified viral RNA dependent RNA Polymerase (RdRp) as a potential drug target due to its significant role in viral replication and transcription. Since no crystal structure is available yet, we report the structural elucidation of CCHFV-RdRp by in-depth homology modeling. Even with low sequence identity, the generated model suggests a similar overall structure as previously reported RdRps. More specifically, the model suggests the presence of structural/functional conserved RdRp motifs for polymerase function, the configuration of uniform spatial arrangement of core RdRp sub-domains, and predicted positively charged entry/exit tunnels, as seen in sNSV polymerases. Extensive pharmacophore modeling based on per-residue energy contribution with investigational drugs allowed the concise mapping of pharmacophoric features and identified potential hits. The combination of pharmacophoric features with interaction energy analysis revealed functionally important residues in the conserved motifs together with in silico predicted common inhibitory binding modes with highly potent reference compounds.
The Crimean-Congo hemorrhagic fever (CCHF) is a devastating viral infection with an extremely high case-to-fatality ratio (ranging from 5–30%, and 50–80% during epidemic events)1 caused by a tick-borne Crimean-Congo hemorrhagic fever virus (CCHFV) from the genus Nairovirus (Family Orthonairovirus, order Bunyaviridae)2,3,4. Being a classically tick-borne disease, the genotypic classification of CCHFV yielded supportive epidemiological data classifying the Hyalomma tick as an important carrier5. Messina et al.6 analyzed human CCHF occurrence data up to 2015 (containing 1,721 geo-positioned occurrences) and reported a full description of its zoonotic niche and the immediate risk to humans. With 10 to 50% mortality and widespread distribution, this arbovirus becomes a substantial concern in case of human-to-human transmission7,8. CCHFV infection starts with non-specific symptoms (from rapid onset high-grade fever, fatigue, myalgia following vomiting and diarrhea) and progresses to a severe disease state with gastrointestinal and cerebral hemorrhages2,9,10. As the disease progresses, CCHFV is observed in the spleen, pulmonary, cardiac, and intestinal tissues in fatal cases in humans11.
The CCHFV viral genome is composed of three single-stranded negative-sense RNA segments referred to as small (S) (~1.6 kb), medium (M) (~5.4 kb) and large (L) (~12.1 kb) segments which encode viral nucleoprotein (NP), the glycoprotein precursor (GPC) and L protein respectively, all of which have complementary 3′ and 5′ termini12. The L protein fragment contains the viral RNA-dependent RNA polymerase (RdRp) domain. Together with viral RNA (vRNA), NP and RdRp form the genomic ribonucleoprotein complexes (RNPs)13, CCHFV viral entry, transcription/replication cycle, glycoprotein maturation and viral assembly have been recently and extensively reviewed10,14,15. In short, CCHFV enters host cells by endocytosis after engaging cell surface receptors using mature glycoproteins (Gn and Gc) and releases the genome. After membrane fusion, the genomic segments are uncoated and transcribed by the L protein into viral mRNA and subsequently translated into NP and L proteins. The endoplasmic reticulum-associated ribosomes translate the GPC. Fragments of newly synthesized NP and L protein are used for replication of genomic RNA, forming RNP. Inside ER and Golgi bodies, the newly CCHFV particles undergo processing and maturation, followed by virion release in Golgi-derived vesicles via exocytosis10,16.
Nairovirus RdRps species share a characteristic right-handed structure with three subdomains (finger, palm, and thumb) indicative and essential for catalytic activity17,18. In all Nairovirus, the RdRp domain maintains six characteristic conserved motifs including PreA/F, A, B, C, D, and E in the central region19,20,21,22. Most of these motifs are located in the palm subdomain and define the formation of the active site21.
Despite a high mortality rate, no licensed antiviral drugs or vaccines are currently available for CCHF23. Efforts have been made in vaccine development24,25 with Bulgarian vaccine26, DNA vaccine27 and the recent NP-based vaccine28. Moreover, a large multi-national consortium, CCHFVaccine, is supporting the development of a CCHFV vaccine which would be a major tool to limit an outbreak (http://www.cchfvaccine.eu/). Broad-spectrum ribavirin has demonstrated antiviral activity when administered in early infection29,30 and also indicated in vitro efficacy in infant mouse models31,32. Favipiravir (T-705) has reported increased activity compared to ribavirin33 with additional activity against two distinct strains of CCHF virus in mouse models34. A screen of candidate nucleoside analog compounds identified 2′-deoxy-2′-fluorocytidine as more potent than ribavirin (200-fold) and T-705 (17-fold)35 and has very recently been reported to be a broad-spectrum inhibitor of Bunyaviruses after in vitro/in vivo studies36. Other experimental therapeutics with no evidence in humans include two repurposed FDA molecules, chloroquine (IC50 39.4 to 28.1 μM) and chlorpromazine (IC50 10.6 to 15.8 μM). MxA (interferon-induced GTPase)37 and ISG20 (interferon-induced exonuclease)23,38 showed direct activity against CCHFV39.
For antiviral drug discovery purposes, druggable viral targets and host proteins are of main interest40,41,42,43,44. CCHFV has a complex genome with multiple proteins involved in processes ranging from virus entry into host cells to viral replication and suppression of the host immune system10. RdRp has a pivotal role in the replication process and is involved in the crucial association of viral RNA to make RNP complexes15.
CCHFV-RdRp core domain of L protein was selected as a target for which no licensed drug has been reported to date. Besides evolutionary conserved motifs in the core region, RdRps have channels/tunnels that connect their catalytic center with the protein surface and emerge as potential targets for developing anti-viral inhibitors45,46,47. The same paradigm is found in the inhibitor design against many lethal viruses such as Japanese encephalitis virus (JEV)48, Zika virus (ZIKV)49,50,51,52, Dengue virus (DENV)53,54,55, West nile virus (WNV)56, HCV57,58,59 and most of drugs targeting Ebola polymerase L (EBOV)60.
Despite the profound antiviral activity of broad-spectrum antivirals activity against CCHFV31,32,33,34,35, the mode of action of T-705 and ribavirin remains suggestive. In accordance with other negative strand viruses, studies showed inhibition of viral RNA polymerases by incorporation of T-705-ribofuranosyl-5′-triphosphate61 and ribavirin-triphosphate62 into nascent RNA strands causing lethal mutagenesis63,64.
Computational methods provide both an alternative and a supplement to tiresome high-throughput screening65,66,67,68, which gave researchers the opportunity to hasten, facilitate and innovate the effectiveness of the overall drug discovery process69,70,71,72,73. Integrated virtual screening methods, including structure-based (SBVS) and ligand-based virtual screening (LBVS), have identified active antiviral compounds against viral epidemics such as influenza virus74, EBOV41,75,76 DENV77,78,79,80 and ZIKV52,81, while others reported the aid of molecular dynamics (MD) simulations and binding free energy calculations in search for potent antivirals50,78,82,83,84,85,86,87,88 and investigating drug resistance mechanisms82,86,89,90,91.
The crystal structure of CCHFV-RdRp has not yet been made available and can be made possible through ab initio methods92,93. Recent advances in homology modeling have proven their effectiveness as an alternative94,95 with retrospective analysis validating the usefulness of homology modeling in SBVS96,97,98,99. Here, we report an optimized cost and time efficient strategy starting from an extensively refined homology model of CCHFV RdRp as a potential druggable target, followed by step-wise pharmacophore-based virtual screening and all-atom backbone molecular dynamics simulation of potential hits.
Materials and Methods
Homology modeling and refinement using Molecular Dynamics (MD) simulations
Homology modeling plays a significant role in the drug discovery process94 with current efforts resulting in models with unprecedented accuracy95,100 even with low sequence identity to the template101,102,103,104,105. Because of the absence of a crystal structural of CCHFV-RdRp, homology modeling was necessary for target structure elucidation. To implement this, CCHFV-RdRp protein sequence was retrieved from the RefSeq database (NCBI Reference Sequence YP_325663). A PSI-BLAST106 search resulted in templates with less than 20% sequence identity with the target of interest. Therefore, an extensive comparative homology modeling protocol was applied in four steps as follows: (1) prior to homology modeling, careful consideration was taken into account for biologically suggestive template selection. For this, various software with different built-in algorithms were used to identify all possible templates including HHpred107, LOMETS (LOcal MEta-Threading-Server)108,MUSTER (MUlti-Source ThreadER)109, ITASSER (Iterative Threading ASSEmbly Refinement)110, RaptorX111 and SWISS MODEL112. Combination of different features of these tools provides an unbiased set-up for template identification.
(2) The models generated from identified templates were structurally compared with respective templates using TM-align113, FATCAT114 and MATRAS115 to identify similarities and differences which result in template selection used to model structural elements. TM-align evaluates optimal structural similarity by TM-score (score > 0.5 indicates the two structures likely to have the same fold in SCOP/CATH). MATRAS compares secondary structural elements and classifies structures based on SCOP superfamily and FOLD. FATCAT allows flexible protein structure comparison to evaluate structural similarity by P-value (structure pairs with P-value < 0.05 indicate significantly similar). Based on several structural attributes between template and model, including comparable secondary structural elements, presence of similar folds, RMSD of equivalence residues, C-alpha backbone RMSD, best models were selected together with the most fitted template.
(3) The final model was built using a restrained-based approach in MODELLER.v9.1101 with the most-fitted template, together with the secondary structural information obtained by manual curation after superimposition between all generated models (step 2) and template. The extracted spatial secondary structure restraints were implemented to model the final structure using the secondary structure module of MODELLER. Target-template alignment was adjusted to reduce the number of misaligned residues.
(4) The models selected in step 2 and 3 were refined by 20 ns MD simulation run and evaluated (before/after MD) via Molprobity (MP) metrics ranked by CASP. Based on the MP-score, the final model was selected, followed by extensive 100 ns MD simulation to check the stability of all backbone atoms. A large ensemble of 1000 snapshots generated by an MD trajectory was used to select the representative conformation from a largest cluster based on RMSD cut-off of ~1 Å. This representative conformation CCHFV-RdRp was further selected to perform modeling studies.
The AMBER16 package using the AMBER ff99SB force field116 was used for unrestrained molecular dynamics simulations. After a stepwise energy minimization and equilibration protocol (as described in a previous studies40,117), the solvated system with explicit TIP3 water molecules was submitted to a production run of 100 ns at constant temperature (300 K) and pressure (1 bar) (detailed protocol in supplementary information).
The MD simulation complete trajectory was collected with a time interval of 2 ps and analyzed using the CPPTRAJ module118. Backbone dynamics were analyzed for both reference template and generated model over a period of 100 ns The most representative conformation of model was obtained as explained in step 4. Later the modeled structure was evaluated through Ramachandran plots generated by MolProbity.
Binding site analysis
A variety of diseases have been associated with the mechanism of involved proteins and their interaction with small molecules119. Identification and characterization of active sites of target proteins using in silico methods are of prime focus to researchers these days41,120. For the identification of active sites, analytical tools are currently being applied since the number of known protein structures is growing fast121,122,123,124,125. To further support the identification of the active site and its residues, COACH meta-server126 and 3DLigSite127 were used. In these methods, the protein function database with ligand-binding templates, BioLiP, is used128. Additionally, the complete protocol for COACH combines the results of several other programs including COFACTOR, Con-Cavity, and FINDSITE resulting in the generation of highly accurate protein-ligand binding site predictions. 3DLigSite incorporates ligand-bounded complexes similar to the query and superimposes these onto the model to predict the binding site.
Molecular docking of potent inhibitors
Small molecules including ribavirin, arbidol and T-705 were selected to be docked at the predicted binding site of CCHFV-RdRp. All these inhibitors were documented to exhibit CCHFV inhibition31,129,130 in vitro and in vivo within a more broad range of RNA viruses131,132,133,134. The optimized and refined CCHFV-RdRp model was minimized and prepared according to the standard protocol of AutoDock (AD) Vina as described elsewhere41,43. Docking grid was set up around the predicted binding site residues covering the conserved structural motifs to reduce the search space for ligand optimization135. The antechamber module from AMBER16116 was used to generate atomic partial charges for the test compounds. The docked complexes were further subjected to MD simulations over a period of 10 ns using the protocol above. As a proof of concept, ribavirin 5′-triphosphate (RTP) and the recently identified 2′-Fluoro-2′-deoxycytidine (2′-FdC) were docked as reference compounds35,36. RTP is the active form of ribavirin136 and 2′-FdC was reported to be 200-fold more potent compared to ribavirin and 17-fold more potent than T-705 against CCHFV,
MM/GBSA binding free energy calculations and Per-residue decomposition analysis
The Molecular Mechanics/Generalized Born Solvent Area (MM/GBSA) method implemented in AMBER 16.0 was employed to estimate the binding affinity of the test compounds. To gain rational insights into the different binding modes, an energy decomposition analysis was performed examining the energetic contribution of each residue interacting with the different compounds. The MM/GBSA approach is well documented in binding free energy calculations137 for antiviral inhibitors138,139.
Pharmacophore model generation and Virtual screening workflow
The reported inhibitors ribavirin, arbidol and T-705 were simulated over a period of 10 ns at the predicted binding site of CCHFV-RdRp to get the most stable conformations of bound ligands. Subsequently, virtual screening process was carried out in several steps starting with the generation of a structure-based pharmacophore model and library generation. For this, the pharmacophore residues were selected based on the energy decomposition profile with reported anti-viral drugs. The constructed model was then added to Ligandscout, to screen the mcule database140 and subjected to the Lipinski’s Rule of Five criterium141. Following this, the generated compound library was further reduced through pharmacokinetics (PK) and pharmacodynamics (PD) filters. Subsequently, drug safety was evaluated through series of PAINS (Pan Assay Interference Compounds) filters142 and Brenk structural alert143 by detecting unwanted toxicophores iv) The screened compounds were then finally docked into the explicit binding site of the refined and optimized CCHFV-RdRp using AD Vina. After screening, the ligand with highest binding affinity was selected for further processing. v) The apo conformation of CCHFV-RdRp was then simulated for a period of 10 ns and four snapshots were taken every 2.5 ns (2.5 ns, 5 ns, 7.5 ns and 10 ns), thereby taking the target flexibility into account144. To finally dilute the compound library, the screened compounds from step (iv) were docked in these four time-dependent conformations. Hits were selected based on binding energy convergence between all four time-dependent conformations of CCHFV-RdRp, and further subjected to 10 ns MD simulations followed by per-residue energy decomposition analysis, as described in molecular dynamics simulation protocol section.
Structure prediction of CCHFV-RdRp
A wide range of viruses encodes the RdRps and have a vital role in replication and transcription of vRNA15. All crystallographic resolved RdRps structures display a similar right hand-like structure with three representative subdomains: finger, palm, and thumb subdomains17. Before structure prediction, high-resolution RdRp structures were examined from segmented negative single-stranded RNA viruses (−ssRNA, order Bunyaviridae) including La Crosse orthobunyavirus (LaCV) (PDB ID: 5amr, 5amq) and Influenza A and B (4wsb, 4wrt) (extensively reviewed elsewhere15,145). These findings revealed the similar overall architecture of L-protein and conserve RdRp core including the conserved finger, palm and thumb subdomains with conserved six structural motifs which are essential for polymerase function47 (Supplementary Fig. S1). Despite the low sequence similarity (~19% identity, ~33% similarity) outside the endonuclease domain146 and conserved structural motifs20, the recently resolved LaCV L- protein fragment comprising 1750 residues (77% of 2263 residues referred to as L1750) revealed a conserved overall architecture with a hetero-trimeric complex of Influenza B virus (formed by the PA, PB1, and PB2 subunits). The RdRp active site is connected with four positively charged tunnels which maintain nucleotide triphosphates entry, nascent strand exit, template entry and template exit, surrounded by conserved residues in orthobunyavirus polymerases21,147. The multiple sequence alignment of the central region of RdRp of Bunyaviridae genera (Fig. S1) highlighted all motifs present in the fingers and palm subdomain to be well conserved including premotif A/F (KxQx5R…K/Rx6E), motif A (Dx2KW), B (QGx5SS), H (KELIL in LaCV), signature motif C (SDD), D (KKT in LaCV), E (Ex2SxF). This pinpoints to the potential of exploiting this evolutionary conserved RdRp core architecture with other related segmented (−) ssRNA viruses to predict their functional and structural features even with a low sequence homology.
Following this, possible templates were identified against CCHFV-RdRp core domain of L-protein using the HHpred server. As expected, HHpred indicated less than 20% identity with the top hits, which included RdRps of LaCV (5AMR) and Influenza B (4WSB, 4WRT) and C virus (5D98).
HHpred uses HHblits, profile-profile and HMM-HMM method, which are able to detect the homologous relationship in evolutionarily related proteins with less than 20% identity, based on conserved structural elements148. All template hits against CCHFV-RdRp revealed homologous to each other (i.e., members of same SCOPe superfamily) and classified in the same class of (−) ssRNA (segmented) viruses including Bunyaviridae and Orthomyxoviridae.
A different set of input parameters, from local to global alignment and HHblits to PSI-BLAST revealed the same results and therefore suggested CCHFV-RdRp to be globally homologous to the putative template hits. Together with an estimated probability of hits (exactly 100% query cover), significant E-value (ranged from 9.8 e−47 to 2.8 e−35 between all RdRps hits) and presence of conserved structural motifs having similar folds suggested these hits as potential templates for homology modeling of CCHFV-RdRp. To rationalize the modeling approach in an unbiased manner, comparative homology modeling protocols were employed using I-TASSER, SWISS-MODEL, LOMETS, RAPTORX and MUSTER which estimate the biological connection of query sequence without the information of any template. Strikingly, all these servers predicted RdRp domains of L1750 (5AMQ) and Influenza A, B and C (4WRT, 4WSB, 5D98) viruses as best templates and recognized identities between 14–20% with a query coverage up to 100% (Supplementary data - Tables S1, S2). The final template-guided model was build using MODELLER (PDB ID: 5 amq, 4wrt, and 4wsb as templates) by taking into account the structural features obtained from the superimposition of modeled structures with the most fitted template. Structural alignment with respective templates using TM-align, FATCAT and MATRAS servers indicated 5amq.A as most favorable template (Supplementary data). The best models from all servers were refined through 20 ns MD simulations and evaluated through Molprobity MP-score (before and after 20 ns production run) tabulated in Table 1. Based on the MP-score, the best CCHFV-RdRp model was generated through MODELLER (MP-score after refinement: 1.44 with 99th percentile) which was further refined through 100 ns MD simulations to optimize overall geometry and to remove clashes in geometry for later analysis.
Overall homology model of CCHFV-RdRp
The overall homology model of CCHFV-RdRp (residues 2043–2994; renumbered 1–951 in homology model) was superimposed on L1750 protein (5AMQ)21 and compared the RdRp core subdomains. The central PB1-like RdRp region of L1750 (residues 758–1433), was previously determined in Influenza polymerase (5AMQ)147. The modelled CCHFV-RdRp core region includes the fingers (overlapping residues 85–430), palm (overlapping residues 265–545) and thumb (residues 546–782) subdomains with conserved structural motifs exposed in the internal core RNA synthesis cavity and display the strikingly similar overall conformation of three core subdomains (Fig. 1A,B) despite the lack of significant sequence homology. Sequence homology with L1750 shows the endonuclease domain of CCHFV-RdRp to be located approximately ~1500 amino acids upstream from the finger domain However, model superimposition represented high structural similarity with the larger α-helical core lobe of L1750 (starting from ~ 85 aa upstream from the finger domain of CCHFV-RdRp) (Fig. 1A). The superimposition of CCHFV-RdRp homology model and L1750 revealed overall structural similarity, with RMSD values of modelled secondary structural elements (Cα-backbone atoms) of 1.32 Å, 1.41 Å and 1.79 Å for finger, palm and thumb subdomains respectively. Other conserved features include the C-terminal Bridge (a helical bundle representation from residues 701–782) which closes the circular architecture of the polymerase core around the internal RNA synthesis chamber, analogous to L175021 (Fig. 1A,B). This characterized the similar conformations of important secondary structural elements such as a partially ordered “α-ribbon” (residues 847–905 in L1750) in the finger domain, California insertion (residues 1021–1044 in L1750) in palm domain (Fig. 1B), a unique exposed helical structure in orthobunyavirus21 and priming loop (residues 1402–1422 in L1750) at the C-terminal end of thumb domain. Reverse template comparison versus structures in the protein databank (PDB) was performed by submitting modelled structure on profunc server149 to further support the structural similarity. This server uses Jess, a fast and accurate 3D-structure search algorithm, to scan auto-generated templates from the query structure against representative structures in PDB. As expected, the reverse template search resulted in RdRp domains of L1750 (5AMQ) and Influenza B (4WRT) among the best hits in terms of E-value (4.48 e−09, 1.77 e−08) and structural similarity (87.9%, 81.8%) (Table S3). The RMSD trajectory plots represent the overall-backbone structure stability of core RdRp subdomains of CCHFV (residues 85–951, in green) compared to L1750 structure (residues 758–1745, in orange) including the central PB1-like RdRp region of L1750 (residues 758–1433) and C-terminal Bridge and thumb ring domains (Fig. 1D). CCHFV-RdRp displayed a similar RMSD value (~3 Å) with the template. The RMSD trajectory of L1750 (residues 758–1745) remained converged from 15 to 100 ns, while backbone Cα-RMSD of CCHFV-RdRp fluctuated in the start and later converged. The slightly higher RMSD in the modelled structure, was presumably caused by protein expansion during simulation to attain a more stable conformation. To investigate the mobility of individual residues, root-mean-square-fluctuations (RMSF) were calculated as shown in Fig. 1E. RMSF plot of Cα-backbone atoms of CCHFV-RdRp model was analyzed together with L1750 to observe the flexibility in the overall structure. Notably, RMSFs of individual RdRp core domains, including fingers, palm, and thumb revealed a similar trend compared to L1750, especially in regions with conserved motifs. Small C-terminal peaks indicated a more open structure due to partially modeled thumb-ring domain of CCHFV-RdRp as compared to L1750 (Fig. 1D). The final CCHFV-RdRp model was selected as a representative conformation obtained from the largest cluster, based on RMSD cut-off of ~1 Å, and evaluated with Molprobity. This indicated 92.12% (876/951 residues) of all residues in Ramachandran favored regions and 97.79% (929/951 residues) of all residues in allowed regions with only 1.78% (17 residues) outliers (Fig. S2).
Potential active site prediction of CCHFV-RdRp
The active site cavity of CCHFV-RdRp and conserved polymerase motifs A-F accurately mimicked the L1750 core subdomains (Fig. 1C). To identify potential binding site residues, COACH and 3DLigandSite were used, together with multiple sequence alignment with closely-related RdRp protein sequences150,151. Both servers predicted Leu234, Arg238, Asp316, Lys319, Trp320, Gln431, Gly432, Ser474, Asp475, Glu533, Phe534, Ser536 and Phe538 as common binding site residues which were shown to be conserved by the multiple sequence alignment (Fig. S1). Interestingly, the structural superimposition with L1750 demonstrated all predicted binding site residues to be exclusively inside the functionally important conserved polymerase motifs of CCHFV-RdRp (Fig. 2A). These include premotif A/F (230-KAQLGGAR), motif A (316-DNTKW) with divalent cation binding Asp316 (Asp1060 in L1750), motif B (431-QGIHHATSS), catalytic signature motif C (474-SDD), and motif E (533-EFYSEF) (binding site residues in bold) (Fig. 2B). More importantly, the predicted binding site residues were found to be stable within ~1 Å fluctuation, as compared to L1750 (Fig. 1B).
The modeling reliability of CCHFV-RdRp allows the prediction of entrance (template and NTP entry channel) and exit (template and nascent strand exit channel) tunnels, calculated with MOLE 2.0152, similar to that reported for L175021 and influenza polymerase147. Figure 2C highlights the prediction of four positively charged tunnels that join together in the active site RNA synthesis chamber where conserved polymerase motifs arbitrate the template-directed RNA synthesis. The template channel entrance in the modelled structure could not be properly defined due to the unavailability of modelled viral RNA binding lobe and α-ribbon albeit minor template entrance channel similarity to L1750 (Fig. 2D). The NTP entrance tunnel was geometrically predicted accurately and aligned with some conserved residues as described for L1750 including Lys230, Arg237 (Lys956, Arg958 in L1750) (motif pre-A/F), Lys319 (Lys1063 in L1750) (motif A), and Lys525 (Lys1228 in L1750) (motif D) (Fig. 2C,D). The product exit tunnel is enclosed by the thumb ring (predicted from 782–951), finger and palm at opposite sides of NTP entry channel as described for L1750. The predicted template exit tunnel is surrounded by thumb, thumb ring and bridge (predicted from 701–782) domains and lined by conserved residues as reported for L1750 including Lys771, Arg772 (Lys1492, Arg1493 in L1750) of bridge domain, and Lys941 and Arg946 (Lys1686, Arg1690 in L1750) of thumb ring (Fig. 2C,D). With these assumptions, the overall structural arrangement of CCHFV-RdRp tunnels and active site prediction are in accordance with the described RNA synthesis in L175021.
Per-residue energy distribution-based pharmacophore model
The structural moieties of binding site of CCHFV-RdRp along with the chemical features of reported compounds (ribavirin, arbidol and T-705) were taken into consideration while creating the guiding pharmacophore model31,32,34,130. For this, a 10 ns MD simulation was carried out on docked ligand-RdRp complexes, followed by MMGBSA and per-residue energy decomposition analysis. This method resulted in enhanced pharmacophore modeling based on the highly contributing residues, and thus construction of a concise subset of small compounds for further selection. All three compounds after 10 ns were found in favorable conformations inside the predicted active site configured by all six structurally conserved RdRp motifs (Fig. 3A). RMSD plots revealed consistent all-atom backbone stability and all three drugs remained inside the binding pocket throughout 10 ns time period except T-705 (orange trajectory), for which small fluctuations were observed (Fig. 3B). The RMSF for all three complexes agrees with lower fluctuations in binding site residues, especially in the palm subdomain which configured all conserved motifs except motif F (Fig. 3B). The pharmacophoric features of all three compounds are displayed in Fig. 3C. 2D interaction plots of average structures after 10 ns are displayed in Fig. 3D. Per-residues decomposition analysis exhibited similar interaction quantification,calculated with arbidol and T-705 complexed with RdRp respectively. Ser536 (−1.917, −1.86 kcal/mol), Phe534 (−0.636, −0.641 kcal/mol), Glu533 (−1.66. −1.515 kcal/mol), Asp316 (−0.825, −0.863 kcal/mol) and Asn317 (−0.736, −0.874 kcal/mol) were found to be top contributing residues (Fig. 3E). Additionally, Gly315 (−1.542 kcal/mol) and Ser474 (−1.482 kcal/mol) also contributed significantly with T-705. Among these highly contributing residues, strong H-bonds were found between nitrogen backbone atom of Ser536 with arbidol (2.38 Å) and T-705 (2.13 Å) respectively. Apart from this, Asn317 forms a H-bond with arbidol (2.81 Å), while energetically favorable residues, Ser474, Asp475 and Glu533 form H-bonds with T-705 (~3 Å). Other selected residues were involved in hydrophobic interactions (Fig. 3C–E).
For RdRp complexed with ribavirin, Ser536 (−3.118 kcal/mol), Ser474 (−1.434 kcal/mol) and Asp475 (−0.637 kcal/mol) are forming H-bonds while other highly contributing residues like Tyr535 (−1.076 kcal/mol), Asn317 (−1.332 kcal/mol) and Leu234 (−0.848 kcal/mol) are mainly involved in hydrophobic interactions (Fig. 3E). Each pharmacophoric feature was used to screen a database with the addition of exclusion volume spheres (as highlighted black) to further narrow the search space for individual compounds.
Virtual screening workflow to identify putative inhibitors
The library generated after scanning the Mcule database with per-residue energy decomposition based pharmacophoric features were subjected to in silico predictions of pharmacokinetics (PK), drug-likeness, toxicity potential, and medicinal chemistry friendliness. This subsequent screening removed substantial hits based as follows: i) Poor ADMET (absorption, distribution, metabolism, excretion, and toxicity) profile as predicted by admetSAR153, ORISIS Property Explorer154 and PreADMET (https://preadmet.bmdrc.kr/). Many hits failed to show desired ADMET characteristics155 due to inhibitory effects on the renal organic cation transporter (ROCT) and to CYP450 enzymes 1A2, 2C9, 2D6, 2C19, and 3A4, together with the prediction to be toxic and carcinogenic. ii) Poor drug-likeness was evaluated by Ghose156, Veber (GSK)157, Pfizer141, Egan (Pharmacia)158, Muegge (Bayer)159 filters and Abbott bioavailability score160. A number of compounds failed to cross these filters devised by established pharmaceutical companies directing the selection of the best molecules for experimental testing.
By these criteria, most virtual hits were found to contain high-risk chemical groups, including epoxides that potentially disturb signal transduction cascades by forming protein adducts161 and quinones that lead to severe oxidative stress through the formation of reactive oxygen species (ROS)162. This extensive step-wise screening resulted in 337 virtual hits after which each compound was docked into the explicit RdRp binding site as predicted by COACH and 3DLigSite. Subsequently, virtual hits were ranked based on binding affinity estimated by an empirical scoring function of AD Vina163. To determine whether the compound is known to be a non-specific aggregator, the compounds were further analyzed using aggregator advisor164. Few compounds were found to have a highly similar scaffold (>78%), previously reported as potential aggregators, leaving 17 hits for further analysis (Table S4). Based on the molecular interactions, binding affinities, and ensemble-based docking, the top-ranked three complexes were selected, and docking conformations were visually analyzed (Fig. S3, Detailed ADMET profile is tabulated in Table S5). In order to identify similar chemical scaffolds of these compounds, a 2D similarity-based was performed using SEA (Similarity Ensemble Approach) which relates proteins (from ChEMBL database) based on the set-wise chemical similarities that exceed a certain threshold (Tanimoto similarity coefficient) among their ligands165. Based on the results, none of the compounds show structural similarity with any known anti-viral inhibitors. Additionally, ChemMapper also predicted the consensus results, which evaluated the 3D-based similarity166. To predict the unfavorable side-effects due to off-target interactions/effects of known molecules and drugs, SwissTargetPrediction web server was used which combines different measures of chemical similarity based on both chemical structure (2D) and molecular shape (3D)167. All 3 compounds were found to have less than 0.5 off-target probability based on cross-validation analysis in ChEMBL for human protein ligands (Table S6).
MD simulations and binding free energy calculations
The RdRp-cmd1, RdRp-cmd2 and RdRp-cmd3, together with RTP, and Fluoro-2′-deoxycytidine (2′-FdC) as two reference compounds, were subjected to 10 ns MD simulations to assess the protein dynamic stability and to analyze decomposed energy contributions of each simulated compound in complex with RdRP. The hydrogen-bond analysis in time, three-dimensional interaction analysis and MSA provided valuable comprehension on the identification of catalytic or inhibitory regions within the CCHFV-RdRp predicted binding site.
CCHFV-RdRp-cmd1/2/3 complex analysis
All simulated compounds in complex with CCHFV-RdRp are displayed in Fig. 4A in proximity to all conserved motifs with stable RMSD trajectories converging within a range of 0.5 Å (Fig. 4B). The complexes were further subjected to root-mean-squared-fluctuation (RMSF) analysis which did not indicate regions of extreme fluctuation. Complexes were found to have a stable network of molecular interactions when analyzed through LigPlot (Fig. 4C), per residue decomposition analysis (Fig. 4D) and H-bond occupancy, defined as the percentage of H-bonds throughout the simulated trajectory (Fig. 4E). Total free binding energy shift from 1 ns to 10 ns together with molecular interactions present in docked and simulated conformations tabulated in Table 2.
For RdRp/cmd1, the 2D Ligplot representation shows the importance of residues Arg238, Glu533, Phe534, and Ser536 belong to motif F and E respectively (Fig. 4C). The guanidine sidechain of Arg238 accepted an H-bond from the urea linker. Glu533 and Phe534 accepted an H-bonds from the nitrogen donor in the benzoxazinone moiety which additionally accepted an H-bond from the backbone amide in Ser536. The H-bond trajectory analysis revealed a high occupancy for Arg238, Glu533, and Ser536 with the highest total energy contribution by Ser536 (Fig. 4D,E). Glu537 interacted predominantly electrostatic. However the total interaction energy categorized it as less contributing. Phe534 displayed a lower overall energetic contribution with a high Van-der-Waals (vdw) contributions. This difference could be accounted to solvation effects. Likewise in the generated pharmacophore model, the Ser536 donates an H-bond to the ethyl carboxylate moiety in Arbidol, to the chloropiperazinone moiety in Favipiravir and the ribose moiety in Ribavirin (Fig. 3C), thus representing a common pharmacophoric feature. With a total interaction free energy result of −33.49 kcal/mol after 10 ns simulation and scored similarly in comparison to the other compounds but displays a substantial increase in predicted binding affinity compared to the investigational drugs (Table 2).
The 2D Ligplot representation of RdRp/cmd2 reveals the importance of residues belongs to motif F (Leu234), motif D (Ser474, Asp475) and motif E (Ser536) (Fig. 4C). Leu234 donated an H-bond from its backbone amide to the benzimidazole moiety of cmd2. Asp475 accepted an H-bond from the hydroxyl of the α-methoxyfenol moiety. With the same hydroxyl, Ser474 acted as H-bond donor and acceptor resulting in two stable H-bonds. Ser536 donated an H-bond from its backbone amide to the α-methoxy group in the same moiety. The H-bond trajectory analysis disclosed nearly 100 percent occupancy for Ser536. Ser474 showed very high occupancy with its hydroxyl sidechain and fair occupancy with its backbone carbonyl as acceptor. The per-residue decomposition analysis of cmd2 displayed notable contributions with Ser474, Glu533, and Ser536 (Fig. 4D). Glu533, however not present in ligplot, especially mediated electrostatic interactions throughout the simulation. Both Ser474 and Ser536 contributed strongly to the overall interaction energy of −32.2 kcal/mol which ranked notably higher than investigational drugs (Table 2). Moreover, when comparing the residues with the highest energetic contribution seen in RdRp complexed with Ribavirin, Favipiravir and Arbidol (Fig. 3C), cmd2 displayed the most common features and energetically contributing residues.
The 2D Ligplot representation of cmd3 displays Arg238, Asp316, and Ser536 as important residues of motif F, A and E interacting through H-bond formation (Fig. 4C). Arg238 donated a H-bond from its guanidine sidechain to the oxygen in the dioxapine moiety of cmd3 and Asp316 accepted a H-bond from the nitrogen in the imidazotriazine moiety. While Ser536 donated a H-bond from its backbone amide to oxygen in the 2H-chromene moiety. The H-bond analysis showed the importance of residues Arg238, Asp316, Ser536 as significant contributors to the overall free interaction energy (Fig. 4D,E). Arg238 displays notable H-bond occupancy throughout the simulation, with a strong electrostatic contribution. The H-bond with Ser536 was formed during the simulation with prominent occupancy. Glu533 interaction, as labeled important in the structure based pharmacophore analysis, is only minorly contributing to the overall free energy.
Additionally, the H-bond analysis displays a low occupancy. However, the neighboring residue, Glu537, is showing the same pattern of electrostatic interaction as seen in cmd1 (Fig. 4C) and Ribavirin (Fig. 3C). The overall interaction energy was calculated as −36.9 kcal/mol. The three simulated compounds score similarly, while markedly scoring more favorable interaction energy values as compared to the investigational drugs (Table 2).
RdRp/RTP and 2FdC complexes
Both reference compounds produced overall consensus results compared to top three compounds. The MD simulations shaped the favorable conformation of RTP and 2′FdC inside the predicted binding site (Fig. 4A) and remained stable throughout 10 ns, with no significant fluctuations observed (Fig. 4B). The generated 2D plots of 2′FdC show a conserved network of molecular interactions as seen in three virtual hit compounds (Fig. 4C), where 2′FdC interacted mainly through a network of H-bonds with Ser474, Asp475 and Asp476 in motif D, and Ser536 in motif E (Figure D). Similarly, the RTP established a network of H-bonds especially through triphosphates with residues of motif D (Asp475, Asp476), motif E (Glu533, Ser536) and motif A (Gly315) (Fig. 4C). Per-residue energy contributions from interacting residues were also found to be comparable as calculated in virtual hits (Fig. 4D). Evidently, both reference compounds showed expected results in terms of interactions and contributing (Table 2).
Even though CCHF is a very old and well-recognized disease, little effort has been put to eradicate either the disease or its symptoms. Besides ribavirin, T-705 has proven historically reliable in reducing CCHF viremia and has been found to be the most efficacious drug from a cohort of similar agents when used against a variety of CCHFV strains130,168. Recently, a screen of candidate nucleoside analog compounds identified 2′-deoxy-2′-fluorocytidine with increased activity in CCHFV compared to ribavirin and T-70535 and showed antiviral activity against several unrelated Bunyaviruses36. Additionally, two repurposed FDA molecules chloroquine and chlorpromazine showed direct activity against CCHFV39. However, there are practically no marketed alternatives available so far.
The CCHFV has a complex genome with multiple proteins involved in processes ranging from virus entry into host cells to viral replication and suppression of the host immune system10,169. Among these, CCHFV-RdRp is involved in critical mechanisms in the virus life cycle, which involves replication/transcription of vRNA in the cytoplasm of infected cell. Therefore, this target is considered an important target against CCHFV (extensively reviewed by15,145). All RdRps for which the crystal structures have been resolved represent a similar right-handed architecture including three subdomains named by their positions in space relative to the palm subdomain17, the conserved polymerase motifs (A-F)145,170 including the signature motif C (GDD in +sRNA19 or SDD in −sRNA viruses)20 and the recently discovered motif H21. The recently resolved crystal structure of LaCV (L1750; residues 1–1750)21 revealed the same overall architecture with influenza polymerase147 including the NTP addition chamber (active site), entrance and exit tunnels. This despite the low sequence identity. Because of the unavailability of CCHFV-RdRp crystal structure in the protein database, an extensive approach to model CCHFV-RdRp was undertaken to further identify potential anti-CCHFV compounds using integrated computational methods. The in silico methods provide a direct and scientifically well-funded basis to guide in vitro methods for antiviral drug discovery70,171. Computational drug discovery has proven to accelerate the challenging process of designing and optimizing new drug candidates70,73. Homology modeling has gained momentum and achieved solid progress with numerous studies reporting refinement of modeled structures followed by MD simulation92,172,173,174,175.
The final CCHFV-RdRp model was optimized and refined by first selecting the most fitted template, followed by extensive post-MD analysis. Models generated from various structure predictions programs enabled the selection of L1750 (PDB ID: 5amq) as the most favorable template, scrutinized by various inclusive structural comparisons (Supplementary information). The final model was generated through MODELLER, by incorporating the spatial secondary structure restraints obtained after 3D structural superimposition of all models with the template (L1750). Even with low sequence identity, the generated model revealed a similar overall structure for all RdRp subdomains when compared with L175017,18 the identified bridge domain (residues 702–782) and partially the thumb ring (residues 783 to ~951) (Fig. 1A,B). Although certain differences emerged, including the α-ribbon, unique California insertion and priming loop in respective domains (Fig. 1B) but the structural superimposition of CCHFV-RdRp with L1750 configured the fingers and thumb subdomains exactly alike insimilar to other polymerases17,18 mainly by their positions, relative to palm subdomain (Fig. 1C). Therefore, sequence similarity (especially in the secondary structural elements), suggested structural/functional relationship, possibly conserved motifs corresponding to conserved binding/enzymatic function are a more appropriate measure of query protein relatedness.
Moreover, the CCHFV-RdRp model was formed as determined in L1750 and other (−) ssRNA viruses, including motifs A-E in the palm and motif F (premotif A) in finger subdomains20,21. Like in L1750 and influenza polymerase147, the arrangement of core subdomains defined the formation of the active site chamber (Fig. 2A) which is connected to exterior by four positively charged tunnels (Fig. 2C), template entry and exit channels. The overall backbone stability of CCHFV-RdRp (residues 85–951) which was comparable to residues 758–1745 in L1750 although the higher stability in L1750 was due to the partial modeling of the thumb ring channel in CCHFV-RdRp model which partly defines the template entry21 (Fig. 1D).
The druggable binding site prediction estimated potential residues which reside inside the conserved polymerase motifs (Fig. 2A). When poliovirus elongation complex structure was superimposed, the conserved Asp residues of signature motif C (Asp475 in model) and A (Asp316 in model) also showed to be in close connection with a divalent cation (Mg+2)176,177 (Fig. 2A). The importance of aspartate residues in motif C (SDD or GDD) and motif A (Dx2KW) is evident from mutational studies which revealed altered polymerase activity in several RdRp viruses178,179,180,181,182,183,184. Because of the highly conserved similar fold (Table S3), the presence of structural conserved motifs and configuration of uniform spatial arrangement of core RdRp domains suggests a conserved evolutionary link between RNA polymerase viruses185,186,187.
MD-refined and optimized CCHFV-RdRp structure incorporated >90% residues in Ramachandran favored region with 17 outliers (Fig. S2). Therefore, the model seems reliable for elucidating CCHFV inhibitors through a structure-based virtual screening (SBVS) protocol. Molecular dynamics based pharmacophore model generation with reported drugs (Fig. 3), ribavirin, arbidol and T-70531,131,132, followed by step-wise virtual screening resulted in final hits that were optimized with rigorous post-MD simulations analysis.
For all compounds including RTP and 2′-FdC reference compounds, the molecular dynamics simulation resulted in an energetically favorable interaction, which provided a more clear insight in the binding mode with structurally conserved RdRp motifs in the palm subdomain. The initial docked pose for each compound already revealed a higher binding affinity, in terms of initial AD-Vina Score and overall binding energy calculated by MM/GBSA module in comparison to the investigational drugs (Table 2). Based on pharmacophore model determination using reported drugs (Fig. 3), functionally important aspartates of motif A (Asp316) and motif C (Asp475, Asp476), together with Glu533 and Ser536 of motif E were found as major H-bond donor/acceptor in the pharmacophore analysis (Fig. 3C,D), additionally evident from the interaction network in all virtual hit (Fig. 4C). The importance of aspartates in RdRps as reported in CCHFV184 and various other mutational studies178,179,180,181,182, highly suggesting a conserved binding feature of cmd1, cmd2 and cmd3, along with Glu533 and Ser536. These residues line deep in the predicted binding pocket of RdRp, potentiating it as an anchor for inhibitory molecules.
Cmd2 ranked highest as potential inhibitor, despite slightly lower interaction free energy as compared to the other simulated compounds. Additionally, it showed the highest H-bond occupancy with Ser536 of motif E and Ser474, Asp475 of motif D as compared to the other simulated compounds.
Conclusively, this comprehensive study introduced a multistep computational approach towards the introduction of novel drugs into the area of anti-CCHF therapy. Merging available protocols has proven to be beneficial because of inherent limitations of the individual sub-strategies70,188. By combining all presented results, we believe the predicted homology model is reliable to hypothesize the 3D-conformation of RdRp core subdomains with conserved motifs and pharmacophoric features maintained by investigational drugs. The resulting model allowed the identification of potential compounds and revealed a common predicted inhibitory effect which was supported in silico with the potent 2′FdC and RTP reference compounds. However, further in vitro testing is necessary to evaluate compound efficacy. Moreover, a widely applicable in silico drug design strategy is depicted for target enzymes without crystal structure.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Dokuzoguz, B. et al. Severity scoring index for Crimean-Congo hemorrhagic fever and the impact of ribavirin and corticosteroids on fatality. Clinical infectious diseases 57, 1270–1274 (2013).
Ergönül, Ö. Crimean-Congo haemorrhagic fever. The Lancet Infectious Diseases 6, 203–214, https://doi.org/10.1016/s1473-3099(06)70435-2 (2006).
Deyde, V. M., Khristova, M. L., Rollin, P. E., Ksiazek, T. G. & Nichol, S. T. Crimean-Congo hemorrhagic fever virus genomics and global diversity. J Virol 80, 8834–8842, https://doi.org/10.1128/JVI.00752-06 (2006).
Adams, M. J. et al. Changes to taxonomy and the International Code of Virus Classification and Nomenclature ratified by the International Committee on Taxonomy of Viruses (2017). Archives of Virology 162, 2505–2538, https://doi.org/10.1007/s00705-017-3358-5 (2017).
Burt, F. J., Spencer, D. C., Leman, P. A., Patterson, B. & Swanepoel, R. Investigation of tick-borne viruses as pathogens of humans in South Africa and evidence of Dugbe virus infection in a patient with prolonged thrombocytopenia. Epidemiology and Infection 116, 353–361 (1996).
Messina, J. P. et al. A global compendium of human Crimean-Congo haemorrhagic fever virus occurrence. Scientific data 2, 150016 (2015).
Appannanavar, S. B. & Mishra, B. An update on crimean congo hemorrhagic Fever. J Glob Infect Dis 3, 285–292, https://doi.org/10.4103/0974-777X.83537 (2011).
Sheikh, A. S. et al. Bi-annual surge of Crimean-Congo haemorrhagic fever (CCHF): a five-year experience. Int J Infect Dis 9, 37–42 (2005).
Whitehouse, C. A. Crimean-Congo hemorrhagic fever. Antiviral Res 64, 145–160, https://doi.org/10.1016/j.antiviral.2004.08.001 (2004).
Zivcec, M., Scholte, F., Spiropoulou, C., Spengler, J. & Bergeron, É. Molecular insights into Crimean-Congo hemorrhagic fever virus. Viruses 8, 106 (2016).
Burt, F. J., Swanepoel, R., Shieh, W.-J. & Smith, J. F. Immunohistochemical and in situ localization of Crimean-Congo hemorrhagic fever (CCHF) virus in human tissues and implications for CCHF pathogenesis. Archives of pathology & laboratory medicine 121, 839 (1997).
King, A. M., Lefkowitz, E., Adams, M. J. & Carstens, E. B. Virus taxonomy: ninth report of the International Committee on Taxonomy of Viruses. (2011).
Carter, S. D. et al. Structure, function and evolution of the Crimean-Congo hemorrhagic fever virus nucleocapsid protein. Journal of virology JVI, 01555–01512 (2012).
Bente, D. A. et al. Crimean-Congo hemorrhagic fever: history, epidemiology, pathogenesis, clinical syndrome and genetic diversity. Antiviral research 100, 159–189 (2013).
Sun, Y., Li, J., Gao, G. F., Tien, P. & Liu, W. Bunyavirales ribonucleoproteins: the viral replication and transcription machinery. Critical reviews in microbiology 44, 522–540 (2018).
Elliott, R. M. The bunyaviridae. (Springer Science & Business Media, 2013).
van Dijk, A. A., Makeyev, E. V. & Bamford, D. H. Initiation of viral RNA-dependent RNA polymerization. Journal of general virology 85, 1077–1093 (2004).
Shatskaya, G. & Dmitrieva, T. Structural organization of viral RNA-dependent RNA polymerases. Biochemistry (Moscow) 78, 231–235 (2013).
Poch, O., Sauvaget, I., Delarue, M. & Tordo, N. Identification of four conserved motifs among the RNA‐dependent polymerase encoding elements. The EMBO journal 8, 3867–3874 (1989).
Müller, R., Poch, O., Delarue, M., Bishop, D. & Bouloy, M. Rift Valley fever virus L segment: correction of the sequence and possible functional role of newly identified regions conserved in RNA-dependent polymerases. Journal of General Virology 75, 1345–1352 (1994).
Gerlach, P., Malet, H., Cusack, S. & Reguera, J. Structural insights into bunyavirus replication and its regulation by the vRNA promoter. Cell 161, 1267–1279 (2015).
Kinsella, E. et al. Sequence determination of the Crimean–Congo hemorrhagic fever virus L segment. Virology 321, 23–28 (2004).
Ergonul, O. Treatment of Crimean-Congo hemorrhagic fever. Antiviral research 78, 125–131 (2008).
Buttigieg, K. R. et al. A novel vaccine against Crimean-Congo Haemorrhagic Fever protects 100% of animals against lethal challenge in a mouse model. PloS one 9, e91516 (2014).
Dowall, S. D., Carroll, M. W. & Hewson, R. Development of vaccines against Crimean-Congo haemorrhagic fever virus. Vaccine (2017).
Papa, A., Papadimitriou, E. & Christova, I. The Bulgarian vaccine Crimean-Congo haemorrhagic fever virus strain. Scandinavian journal of infectious diseases 43, 225–229 (2011).
Garrison, A. R. et al. A DNA vaccine for Crimean-Congo hemorrhagic fever protects against disease and death in two lethal mouse models. PLoS neglected tropical diseases 11, e0005908 (2017).
Zivcec, M., Safronetz, D., Scott, D. P., Robertson, S. & Feldmann, H. Nucleocapsid protein-based vaccine provides protection in mice against lethal Crimean-Congo hemorrhagic fever virus challenge. PLoS neglected tropical diseases 12, e0006628 (2018).
Fisgin, N. T., Ergonul, O., Doganci, L. & Tulek, N. The role of ribavirin in the therapy of Crimean-Congo hemorrhagic fever: early use is promising. European journal of clinical microbiology & infectious diseases 28, 929–933 (2009).
Ozbey, S. B., Kader, Ç., Erbay, A. & Ergönül, Ö. Early use of ribavirin is beneficial in Crimean-Congo hemorrhagic fever. Vector-Borne and Zoonotic Diseases 14, 300–302 (2014).
Oestereich, L. et al. Evaluation of antiviral efficacy of ribavirin, arbidol, and T-705 (favipiravir) in a mouse model for Crimean-Congo hemorrhagic fever. PLoS neglected tropical diseases 8, e2804 (2014).
Johnson, S. et al. Ribavirin for treating Crimean Congo haemorrhagic fever. Cochrane Database of Systematic Reviews (2018).
Oestereich, L. et al. Efficacy of favipiravir alone and in combination with ribavirin in a lethal, immunocompetent mouse model of Lassa fever. The Journal of infectious diseases 213, 934–938 (2015).
Hawman, D. W. et al. Favipiravir (T-705) but not ribavirin is effective against two distinct strains of Crimean-Congo hemorrhagic fever virus in mice. Antiviral Research (2018).
Welch, S. R. et al. Identification of 2′-deoxy-2′-fluorocytidine as a potent inhibitor of Crimean-Congo hemorrhagic fever virus replication using a recombinant fluorescent reporter virus. Antiviral research 147, 91–99 (2017).
Smee, D. F., Jung, K.-H., Westover, J. & Gowen, B. B. 2′-Fluoro-2′-deoxycytidine is a broad-spectrum inhibitor of bunyaviruses in vitro and in phleboviral disease mouse models. Antiviral research 160, 48–54 (2018).
Andersson, I. et al. Human MxA protein inhibits the replication of Crimean-Congo hemorrhagic fever virus. J Virol 78, 4323–4329 (2004).
Mirazimi, A. Old and new treatment strategies. Crimean-Congo Hemorrhagic Fever: A global perspective, 258–260 (2007).
Ferraris, O. et al. Evaluation of Crimean-Congo hemorrhagic fever virus in vitro inhibition by chloroquine and chlorpromazine, two FDA approved molecules. Antiviral research 118, 75–81 (2015).
Usman Mirza, M. et al. Towards peptide vaccines against Zika virus: Immunoinformatics combined with molecular dynamics simulations to predict antigenic epitopes of Zika viral proteins. Sci Rep 6, 37313, https://doi.org/10.1038/srep37313 (2016).
Mirza, M. U. & Ikram, N. Integrated Computational Approach for Virtual Hit Identification against Ebola Viral Proteins VP35 and VP40. Int J Mol Sci 17, https://doi.org/10.3390/ijms17111748 (2016).
Bassetto, M., Massarotti, A., Coluccia, A. & Brancale, A. Structural biology in antiviral drug discovery. Curr Opin Pharmacol 30, 116–130, https://doi.org/10.1016/j.coph.2016.08.014 (2016).
Mirza, M. U., Ghori, N. U., Ikram, N., Adil, A. R. & Manzoor, S. Pharmacoinformatics approach for investigation of alternative potential hepatitis C virus nonstructural protein 5B inhibitors. Drug Des Devel Ther 9, 1825–1841, https://doi.org/10.2147/DDDT.S75886 (2015).
Ganesan, A. & Barakat, K. Applications of computer-aided approaches in the development of hepatitis C antiviral agents. Expert Opin Drug Discov 12, 407–425, https://doi.org/10.1080/17460441.2017.1291628 (2017).
Jácome, R., Becerra, A., de León, S. P. & Lazcano, A. Structural analysis of monomeric RNA-dependent polymerases: Evolutionary and therapeutic implications. PloS one 10, e0139001 (2015).
Van Der Linden, L. et al. The RNA template channel of the RNA-dependent RNA polymerase as a target for development of antiviral therapy of multiple genera within a virus family. PLoS pathogens 11, e1004733 (2015).
Ferrer-Orta, C., Arias, A., Escarmís, C. & Verdaguer, N. A comparison of viral RNA-dependent RNA polymerases. Current opinion in structural biology 16, 27–34 (2006).
Lu, G. & Gong, P. Crystal structure of the full-length Japanese encephalitis virus NS5 reveals a conserved methyltransferase-polymerase interface. PLoS pathogens 9, e1003549 (2013).
Godoy, A. S. et al. Crystal structure of Zika virus NS5 RNA-dependent RNA polymerase. Nature communications 8, 14764 (2017).
Zhang, C. et al. Structure of the NS5 methyltransferase from Zika virus and implications in inhibitor design. Biochemical and biophysical research communications 492, 624–630 (2017).
Hercík, K. et al. Adenosine triphosphate analogs can efficiently inhibit the Zika virus RNA-dependent RNA polymerase. Antiviral research 137, 131–133 (2017).
Pattnaik, A. et al. Discovery of a non-nucleoside RNA polymerase inhibitor for blocking Zika virus replication through in silico screening. Antiviral research 151, 78–86 (2018).
Noble, C. G. et al. Conformational flexibility of the Dengue virus RNA-dependent RNA polymerase revealed by a complex with an inhibitor. Journal of virology JVI, 00045–00013 (2013).
Noble, C. G. et al. Strategies for development of dengue virus inhibitors. Antiviral research 85, 450–462 (2010).
El Sahili, A. & Lescar, J. Dengue virus non-structural protein 5. Viruses 9, 91 (2017).
Malet, H. et al. Crystal structure of the RNA polymerase domain of the West Nile virus non-structural protein 5. Journal of Biological Chemistry 282, 10678–10689 (2007).
De Francesco, R., Tomei, L., Altamura, S., Summa, V. & Migliaccio, G. Approaching a new era for hepatitis C virus therapy: inhibitors of the NS3-4A serine protease and the NS5B RNA-dependent RNA polymerase. Antiviral research 58, 1–16 (2003).
Dhanak, D. et al. Identification and biological characterization of heterocyclic inhibitors of the hepatitis C virus RNA-dependent RNA polymerase. Journal of Biological Chemistry 277, 38322–38327 (2002).
Gemma, S. et al. HCV-targeted antivirals: current status and future challenges. Current pharmaceutical design 20, 3445–3464 (2014).
Mirza, M. U. et al. Perspectives towards antiviral drug discovery against Ebola virus. Journal of medical virology.
Jin, Z., Smith, L. K., Rajwanshi, V. K., Kim, B. & Deval, J. The ambiguous base-pairing and high substrate efficiency of T-705 (favipiravir) ribofuranosyl 5′-triphosphate towards influenza A virus polymerase. PloS one 8, e68347 (2013).
Eriksson, B. et al. Inhibition of influenza virus ribonucleic acid polymerase by ribavirin triphosphate. Antimicrobial agents and chemotherapy 11, 946–951 (1977).
Baranovich, T. et al. T-705 (favipiravir) induces lethal mutagenesis in influenza A H1N1 viruses in vitro. Journal of virology JVI, 02346–02312 (2013).
Crotty, S. et al. The broad-spectrum antiviral ribonucleoside ribavirin is an RNA virus mutagen. Nature medicine 6, 1375 (2000).
Blundell, T. L., Jhoti, H. & Abell, C. High-throughput crystallography for lead discovery in drug design. Nat Rev Drug Discov 1, 45–54, https://doi.org/10.1038/nrd706 (2002).
Cavasotto, C. N., Orry, W. & Andrew, J. Ligand docking and structure-based virtual screening in drug discovery. Current topics in medicinal chemistry 7, 1006–1014 (2007).
Cheng, T., Li, Q., Zhou, Z., Wang, Y. & Bryant, S. H. Structure-based virtual screening for drug discovery: a problem-centric review. AAPS J 14, 133–141, https://doi.org/10.1208/s12248-012-9322-0 (2012).
Drwal, M. N. & Griffith, R. Combination of ligand-and structure-based methods in virtual screening. Drug Discovery Today: Technologies 10, e395–e401 (2013).
Song, C. M., Lim, S. J. & Tong, J. C. Recent advances in computer-aided drug design. Brief Bioinform 10, 579–591, https://doi.org/10.1093/bib/bbp023 (2009).
Hung, C. L. & Chen, C. C. Computational approaches for drug discovery. Drug Dev Res 75, 412–418, https://doi.org/10.1002/ddr.21222 (2014).
Jorgensen, W. L. The many roles of computation in drug discovery. Science 303, 1813–1818, https://doi.org/10.1126/science.1096361 (2004).
Ganesan, A., Coote, M. L. & Barakat, K. Molecular dynamics-driven drug discovery: leaping forward with confidence. Drug Discov Today 22, 249–269, https://doi.org/10.1016/j.drudis.2016.11.001 (2017).
Macalino, S. J., Gosu, V., Hong, S. & Choi, S. Role of computer-aided drug design in modern drug discovery. Arch Pharm Res 38, 1686–1701, https://doi.org/10.1007/s12272-015-0640-5 (2015).
Du, J., Cross, T. A. & Zhou, H.-X. Recent progress in structure-based anti-influenza drug design. Drug discovery today 17, 1111–1120 (2012).
Madrid, P. B. et al. A systematic screen of FDA-approved drugs for inhibitors of biological threat agents. PLoS One 8, e60579, https://doi.org/10.1371/journal.pone.0060579 (2013).
Shurtleff, A. C., Nguyen, T. L., Kingery, D. A. & Bavari, S. Therapeutics for filovirus infection: traditional approaches and progress towards in silico drug design. Expert Opin Drug Discov 7, 935–954, https://doi.org/10.1517/17460441.2012.714364 (2012).
Leela, S. L. et al. Drug repurposing of minocycline against dengue virus infection. Biochem Biophys Res Commun 478, 410–416, https://doi.org/10.1016/j.bbrc.2016.07.029 (2016).
Luzhkov, V., Decroly, E., Canard, B., Selisko, B. & Åqvist, J. Evaluation of adamantane derivatives as inhibitors of dengue virus mRNA cap methyltransferase by docking and molecular dynamics simulations. Molecular informatics 32, 155–164 (2013).
Wang, Q. Y. et al. A small-molecule dengue virus entry inhibitor. Antimicrob Agents Chemother 53, 1823–1831, https://doi.org/10.1128/AAC.01148-08 (2009).
Zhou, Z. et al. Antiviral Compounds Discovered by Virtual Screening of Small− Molecule Libraries against Dengue Virus E Protein. ACS chemical biology 3, 765–775 (2008).
Nitsche, C. In Dengue and Zika: Control and Antiviral Treatment Strategies 175-186 (Springer, 2018).
Hou, T. & Yu, R. Molecular dynamics and free energy studies on the wild-type and double mutant HIV-1 protease complexed with amprenavir and two amprenavir-related inhibitors: mechanism for binding and drug resistance. Journal of medicinal chemistry 50, 1177–1188 (2007).
Tu, J., Li, J. J., Shan, Z. J. & Zhai, H. L. Exploring the binding mechanism of Heteroaryldihydropyrimidines and Hepatitis B Virus capsid combined 3D-QSAR and molecular dynamics. Antiviral research 137, 151–164 (2017).
Anusuya, S. & Gromiha, M. M. Quercetin derivatives as non-nucleoside inhibitors for dengue polymerase: molecular docking, molecular dynamics simulation, and binding free energy calculation. Journal of Biomolecular Structure and Dynamics 35, 2895–2909 (2017).
Guan, S. et al. Exploration of binding and inhibition mechanism of a small molecule inhibitor of influenza virus H1N1 hemagglutinin by molecular dynamics simulation. Scientific Reports 7 (2017).
Bhakat, S., Martin, A. J. & Soliman, M. E. An integrated molecular dynamics, principal component analysis and residue interaction network approach reveals the impact of M184V mutation on HIV reverse transcriptase resistance to lamivudine. Molecular BioSystems 10, 2215–2228 (2014).
Speelman, B., Brooks, B. R. & Post, C. B. Molecular dynamics simulations of human rhinovirus and an antiviral compound. Biophysical Journal 80, 121–129 (2001).
Mottin, M. et al. Molecular dynamics simulations of Zika virus NS3 helicase: Insights into RNA binding site activity. Biochemical and Biophysical Research Communications (2017).
Pan, D. et al. Computational study on the drug resistance mechanism of hepatitis C virus NS5B RNA-dependent RNA polymerase mutants to BMS-791325 by molecular dynamics simulation and binding free energy calculations. Chemometrics and Intelligent Laboratory Systems 154, 185–193 (2016).
Leonis, G., Steinbrecher, T. & Papadopoulos, M. G. A contribution to the drug resistance mechanism of Darunavir, Amprenavir, Indinavir, and Saquinavir complexes with HIV-1 protease due to flap mutation I50V: A systematic MM–PBSA and thermodynamic integration study. Journal of chemical information and modeling 53, 2141–2153 (2013).
Pan, P., Li, L., Li, Y., Li, D. & Hou, T. Insights into susceptibility of antiviral drugs against the E119G mutant of 2009 influenza A (H1N1) neuraminidase by molecular dynamics simulations and free energy calculations. Antiviral research 100, 356–364 (2013).
Fan, H. & Mark, A. E. Refinement of homology-based protein structures by molecular dynamics simulation techniques. Protein Sci 13, 211–220, https://doi.org/10.1110/ps.03381404 (2004).
Sali, A. & Blundell, T. L. Comparative protein modelling by satisfaction of spatial restraints. J Mol Biol 234, 779–815, https://doi.org/10.1006/jmbi.1993.1626 (1993).
Cavasotto, C. N. & Phatak, S. S. Homology modeling in drug discovery: current trends and applications. Drug Discov Today 14, 676–683, https://doi.org/10.1016/j.drudis.2009.04.006 (2009).
Hillisch, A., Pineda, L. F. & Hilgenfeld, R. Utility of homology models in the drug discovery process. Drug Discov Today 9, 659–669, https://doi.org/10.1016/S1359-6446(04)03196-4 (2004).
Oshiro, C. et al. Performance of 3D-database molecular docking studies into homology models. Journal of Medicinal Chemistry 47, 764–767, https://doi.org/10.1021/jm0300781 (2004).
Kairys, V., Fernandes, M. X. & Gilson, M. K. Screening drug-like compounds by docking to homology models: a systematic study. J Chem Inf Model 46, 365–379, https://doi.org/10.1021/ci050238c (2006).
Fernandes, M. X., Kairys, V. & Gilson, M. K. Comparing ligand interactions with multiple receptors via serial docking. J Chem Inf Comput Sci 44, 1961–1970, https://doi.org/10.1021/ci049803m (2004).
McGovern, S. L. & Shoichet, B. K. Information decay in molecular docking screens against holo, apo, and modeled conformations of enzymes. J Med Chem 46, 2895–2907, https://doi.org/10.1021/jm0300330 (2003).
Larsson, P., Wallner, B., Lindahl, E. & Elofsson, A. Using multiple templates to improve quality of homology models in automated homology modeling. Protein Science 17, 990–1002, https://doi.org/10.1110/ps.073344908 (2008).
Fiser, A. & Sali, A. Modeller: generation and refinement of homology-based protein structure models. Methods Enzymol 374, 461–491, https://doi.org/10.1016/S0076-6879(03)74020-8 (2003).
Fan, H. et al. Molecular Docking Screens Using Comparative Models of Proteins. Journal of Chemical Information and Modeling 49, 2512–2527, https://doi.org/10.1021/ci9003706 (2009).
Mariani, V., Kiefer, F., Schmidt, T., Haas, J. & Schwede, T. Assessment of template based protein structure predictions in CASP9. Proteins 79(Suppl 10), 37–58, https://doi.org/10.1002/prot.23177 (2011).
Huang, Y. J., Mao, B., Aramini, J. M. & Montelione, G. T. Assessment of template-based protein structure predictions in CASP10. Proteins 82(Suppl 2), 43–56, https://doi.org/10.1002/prot.24488 (2014).
Webb, B. & Sali, A. Protein Structure Modeling with MODELLER. Protein Structure Prediction, 3rd Edition 1137, 1–15, https://doi.org/10.1007/978-1-4939-0366-5_1 (2014).
Altschul, S. F. et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 25, 3389–3402 (1997).
Söding, J., Biegert, A. & Lupas, A. N. The HHpred interactive server for protein homology detection and structure prediction. Nucleic acids research 33, W244–W248 (2005).
Wu, S. & Zhang, Y. LOMETS: a local meta-threading-server for protein structure prediction. Nucleic acids research 35, 3375–3382 (2007).
Wu, S. & Zhang, Y. MUSTER: improving protein sequence profile–profile alignments by using multiple sources of structure information. Proteins: Structure, Function, and Bioinformatics 72, 547–556 (2008).
Roy, A., Kucukural, A. & Zhang, Y. I-TASSER: a unified platform for automated protein structure and function prediction. Nature protocols 5, 725 (2010).
Källberg, M. et al. Template-based protein structure modeling using the RaptorX web server. Nature protocols 7, 1511 (2012).
Schwede, T., Kopp, J., Guex, N. & Peitsch, M. C. SWISS-MODEL: an automated protein homology-modeling server. Nucleic acids research 31, 3381–3385 (2003).
Zhang, Y. & Skolnick, J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic acids research 33, 2302–2309 (2005).
Ye, Y. & Godzik, A. FATCAT: a web server for flexible structure comparison and structure similarity searching. Nucleic acids research 32, W582–W585 (2004).
Kawabata, T. MATRAS: a program for protein 3D structure comparison. Nucleic acids research 31, 3367–3369 (2003).
Maier, J. A. et al. ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB. J Chem Theory Comput 11, 3696–3713, https://doi.org/10.1021/acs.jctc.5b00255 (2015).
Jabbar, B. et al. Antigenic Peptide Prediction From E6 and E7 Oncoproteins of HPV Types 16 and 18 for Therapeutic Vaccine Design Using Immunoinformatics and MD Simulation Analysis. Frontiers in immunology 9 (2018).
Roe, D. R. & Cheatham, T. E. III. PTRAJ and CPPTRAJ: Software for Processing and Analysis of Molecular Dynamics Trajectory Data. J Chem Theory Comput 9, 3084–3095, https://doi.org/10.1021/ct400341p (2013).
Barabási, A.-L., Gulbahce, N. & Loscalzo, J. Network medicine: a network-based approach to human disease. Nature reviews genetics 12, 56 (2011).
Katsila, T., Spyroulias, G. A., Patrinos, G. P. & Matsoukas, M. T. Computational approaches in target identification and drug discovery. Comput Struct Biotechnol J 14, 177–184, https://doi.org/10.1016/j.csbj.2016.04.004 (2016).
Sotriffer, C. & Klebe, G. Identification and mapping of small-molecule binding sites in proteins: computational tools for structure-based drug design. Farmaco 57, 243–251 (2002).
Yu, J., Zhou, Y., Tanaka, I. & Yao, M. Roll: a new algorithm for the detection of protein pockets and cavities with a rolling probe sphere. Bioinformatics 26, 46–52, https://doi.org/10.1093/bioinformatics/btp599 (2010).
Weisel, M., Proschak, E. & Schneider, G. PocketPicker: analysis of ligand binding-sites with shape descriptors. Chem Cent J 1, 7, https://doi.org/10.1186/1752-153X-1-7 (2007).
Nisius, B., Sha, F. & Gohlke, H. Structure-based computational analysis of protein binding sites for function and druggability prediction. J Biotechnol 159, 123–134, https://doi.org/10.1016/j.jbiotec.2011.12.005 (2012).
Henrich, S. et al. Computational approaches to identifying and characterizing protein binding sites for ligand design. J Mol Recognit 23, 209–219, https://doi.org/10.1002/jmr.984 (2010).
Yang, J. Y., Roy, A. & Zhang, Y. Protein-ligand binding site recognition using complementary binding-specific substructure comparison and sequence profile alignment. Bioinformatics 29, 2588–2595, https://doi.org/10.1093/bioinformatics/btt447 (2013).
Wass, M. N., Kelley, L. A. & Sternberg, M. J. E. 3DLigandSite: predicting ligand-binding sites using similar structures. Nucleic Acids Research 38, W469–W473, https://doi.org/10.1093/nar/gkq406 (2010).
Yang, J., Roy, A. & Zhang, Y. BioLiP: a semi-manually curated database for biologically relevant ligand-protein interactions. Nucleic Acids Res 41, D1096–1103, https://doi.org/10.1093/nar/gks966 (2013).
Johnson, S. et al. Ribavirin for treating Crimean Congo haemorrhagic fever. The Cochrane Library (2017).
Tignor, G. H. & Hanham, C. A. Ribavirin efficacy in an in vivo model of Crimean-Congo hemorrhagic fever virus (CCHF) infection. Antiviral Res 22, 309–325 (1993).
Shi, L. et al. Antiviral activity of arbidol against influenza A virus, respiratory syncytial virus, rhinovirus, coxsackie virus and adenovirus in vitro and in vivo. Archives of virology 152, 1447–1455 (2007).
Brooks, M. J. et al. Antiviral activity of arbidol, a broad‐spectrum drug for use against respiratory viruses, varies according to test conditions. Journal of medical virology 84, 170–181 (2012).
Furuta, Y. et al. In vitro and in vivo activities of anti-influenza virus compound T-705. Antimicrobial agents and chemotherapy 46, 977–981 (2002).
Gowen, B. B. et al. In vitro and in vivo activities of T-705 against arenavirus and bunyavirus infections. Antimicrobial agents and chemotherapy 51, 3168–3176 (2007).
Trott, O. & Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31, 455–461, https://doi.org/10.1002/jcc.21334 (2010).
Patterson, J. L. & Fernandez-Larsson, R. Molecular mechanisms of action of ribavirin. Reviews of infectious diseases 12, 1139–1146 (1990).
Hou, T., Wang, J., Li, Y. & Wang, W. Assessing the performance of the MM/PBSA and MM/GBSA methods. 1. The accuracy of binding free energy calculations based on molecular dynamics simulations. Journal of chemical information and modeling 51, 69–82 (2010).
Srivastava, H. K. & Sastry, G. N. Molecular dynamics investigation on a series of HIV protease inhibitors: assessing the performance of MM-PBSA and MM-GBSA approaches. Journal of chemical information and modeling 52, 3088–3098 (2012).
Tan, J. J., Zu Chen, W. & Wang, C. X. Investigating interactions between HIV-1 gp41 and inhibitors by molecular dynamics simulation and MM–PBSA/GBSA calculations. Journal of Molecular Structure: THEOCHEM 766, 77–82 (2006).
Kiss, R., Szalai, F. & Sandor, M. Mcule.com: A public web service for drug discovery. Abstracts of Papers of the American Chemical Society 243 (2012).
Lipinski, C. A., Lombardo, F., Dominy, B. W. & Feeney, P. J. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 46, 3–26 (2001).
Baell, J. B. & Holloway, G. A. New Substructure Filters for Removal of Pan Assay Interference Compounds (PAINS) from Screening Libraries and for Their Exclusion in Bioassays. Journal of Medicinal Chemistry 53, 2719–2740, https://doi.org/10.1021/jm901137j (2010).
Brenk, R. et al. Lessons learnt from assembling screening libraries for drug discovery for neglected diseases. ChemMedChem 3, 435–444, https://doi.org/10.1002/cmdc.200700139 (2008).
Campbell, A. J., Lamb, M. L. & Joseph-McCarthy, D. Ensemble-based docking using biased molecular dynamics. Journal of chemical information and modeling 54, 2127–2138 (2014).
Amroun, A., Priet, S., de Lamballerie, X. & Quérat, G. Bunyaviridae RdRps: structure, motifs, and RNA synthesis machinery. Critical reviews in microbiology 43, 753–778 (2017).
Reguera, J., Weber, F. & Cusack, S. Bunyaviridae RNA polymerases (L-protein) have an N-terminal, influenza-like endonuclease domain, essential for viral cap-dependent transcription. PLoS pathogens 6, e1001101 (2010).
Pflug, A., Guilligay, D., Reich, S. & Cusack, S. Structure of influenza A polymerase bound to the viral RNA promoter. Nature 516, 355 (2014).
Remmert, M., Biegert, A., Hauser, A. & Söding, J. HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment. Nature methods 9, 173 (2012).
Laskowski, R. A., Watson, J. D. & Thornton, J. M. ProFunc: a server for predicting protein function from 3D structure. Nucleic acids research 33, W89–W93 (2005).
Morgenstern, B., Prohaska, S. J., Pohler, D. & Stadler, P. F. Multiple sequence alignment with user-defined anchor points. Algorithms Mol Biol 1, 6, https://doi.org/10.1186/1748-7188-1-6 (2006).
Bi, C. P. Multiple sequence local alignment using Monte Carlo EM algorithm. Bioinformatics Research and Applications, Proceedings 4463, 465–476 (2007).
Sehnal, D. et al. MOLE 2.0: advanced approach for analysis of biomacromolecular channels. Journal of cheminformatics 5, 39 (2013).
Cheng, F. et al. admetSAR: a comprehensive source and free tool for assessment of chemical ADMET properties. J Chem Inf Model 52, 3099–3105, https://doi.org/10.1021/ci300367a (2012).
Cabrera, M. et al. Synthetic chalcones, flavanones, and flavones as antitumoral agents: Biological evaluation and structure-activity relationships. Bioorganic & Medicinal Chemistry 15, 3356–3367, https://doi.org/10.1016/j.bmc.2007.03.031 (2007).
Moroy, G., Martiny, V. Y., Vayer, P., Villoutreix, B. O. & Miteva, M. A. Toward in silico structure-based ADMET prediction in drug discovery. Drug Discov Today 17, 44–55, https://doi.org/10.1016/j.drudis.2011.10.023 (2012).
Ghose, A. K., Viswanadhan, V. N. & Wendoloski, J. J. A knowledge-based approach in designing combinatorial or medicinal chemistry libraries for drug discovery. 1. A qualitative and quantitative characterization of known drug databases. J Comb Chem 1, 55–68 (1999).
Veber, D. F. et al. Molecular properties that influence the oral bioavailability of drug candidates. Journal of Medicinal Chemistry 45, 2615–2623, https://doi.org/10.1021/jm020017n (2002).
Egan, W. J., Merz, K. M. & Baldwin, J. J. Prediction of drug absorption using multivariate statistics. Journal of Medicinal Chemistry 43, 3867–3877, https://doi.org/10.1021/jm000292e (2000).
Muegge, I., Heald, S. L. & Brittelli, D. Simple selection criteria for drug-like chemical matter. J Med Chem 44, 1841–1846 (2001).
Karplus, M. & Kuriyan, J. Molecular dynamics and protein function. Proceedings of the National Academy of Sciences 102, 6679–6685 (2005).
Blagg, J. In Burger’s Medicinal Chemistry and Drug Discovery 301–334 (2010).
Rishton, G. M. Nonleadlikeness and leadlikeness in biochemical screening. Drug Discov Today 8, 86–96 (2003).
Bohm, D. & Stapp, H. P. The Undivided Universe: An ontological interpretation of Quantum Theory. American Journal of Physics 62, 958–960, https://doi.org/10.1119/1.17695 (1994).
Irwin, J. J. et al. An Aggregation Advisor for Ligand Discovery. J Med Chem 58, 7076–7087, https://doi.org/10.1021/acs.jmedchem.5b01105 (2015).
Keiser, M. J. et al. Relating protein pharmacology by ligand chemistry. Nature biotechnology 25, 197 (2007).
Gong, J. et al. ChemMapper: a versatile web server for exploring pharmacology and chemical structure association based on molecular 3D similarity method. Bioinformatics 29, 1827–1829 (2013).
Gfeller, D. et al. SwissTargetPrediction: a web server for target prediction of bioactive small molecules. Nucleic acids research 42, W32–W38 (2014).
Paragas, J., Whitehouse, C. A., Endy, T. P. & Bray, M. A simple assay for determining antiviral activity against Crimean-Congo hemorrhagic fever virus. Antiviral Res 62, 21–25, https://doi.org/10.1016/j.antiviral.2003.11.006 (2004).
Dowall, S. D. et al. A Crimean-Congo hemorrhagic fever (CCHF) viral vaccine expressing nucleoprotein is immunogenic but fails to confer protection against lethal disease. Human Vaccines & Immunotherapeutics 12, 519–527, https://doi.org/10.1080/21645515.2015.1078045 (2016).
Hussein, I. T., Haseeb, A., Haque, A. & Mir, M. A. In Advances in applied microbiology Vol. 74 35–75 (Elsevier, 2011).
Hartenfeller, M. & Schneider, G. De novo drug design. Chemoinformatics and computational chemical biology, 299–323 (2011).
Chen, J. & Brooks, C. L. Can molecular dynamics simulations provide high-resolution refinement of protein structure? Proteins: Structure, Function, and Bioinformatics 67, 922–930 (2007).
Chopra, G., Summa, C. M. & Levitt, M. Solvent dramatically affects protein structure refinement. Proc Natl Acad Sci USA 105, 20239–20244, https://doi.org/10.1073/pnas.0810818105 (2008).
Ishitani, R., Terada, T. & Shimizu, K. Refinement of comparative models of protein structure by using multicanonical molecular dynamics simulations. Molecular Simulation 34, 327–336, https://doi.org/10.1080/08927020801930539 (2008).
Jagielska, A., Wroblewska, L. & Skolnick, J. Protein model refinement using an optimized physics-based all-atom force field. Proceedings of the National Academy of Sciences of the United States of America 105, 8268–8273, https://doi.org/10.1073/pnas.0800054105 (2008).
O’Reilly, E. K. & Kao, C. C. Analysis of RNA-dependent RNA polymerase structure and function as guided by known polymerase structures and computer predictions of secondary structure. Virology 252, 287–303 (1998).
Rothwell, P. J. & Waksman, G. Structure and mechanism of DNA polymerases. Advances in protein chemistry 71, 401–440 (2005).
Zhou, Y., Zheng, H., Gao, F., Tian, D. & Yuan, S. Mutational analysis of the SDD sequence motif of a PRRSV RNA-dependent RNA polymerase. Science China Life Sciences 54, 870–879 (2011).
Sánchez, A. B. & Juan, C. Genetic and biochemical evidence for an oligomeric structure of the functional L polymerase of the prototypic arenavirus lymphocytic choriomeningitis virus. Journal of virology 79, 7262–7268 (2005).
Arnold, J. J., Ghosh, S. K. B. & Cameron, C. E. Poliovirus RNA-dependent RNA polymerase (3Dpol) divalent cation modulation of primer, template, and nucleotide selection. Journal of Biological Chemistry 274, 37060–37069 (1999).
Vázquez, A. L., Alonso, J. M. M. & Parra, F. Mutation analysis of the GDD sequence motif of a calicivirus RNA-dependent RNA polymerase. Journal of virology 74, 3888–3891 (2000).
Biswas, S. K. & Nayak, D. P. Mutational analysis of the conserved motifs of influenza A virus polymerase basic protein 1. Journal of virology 68, 1819–1826 (1994).
EWAN, E., DUNN, D. C., HONG, J. & RICHARD, M. E. Transcription of a recombinant bunyavirus RNA template by transiently expressed bunyavirus proteins. Virology 211, 133–143 (1995).
Bergeron, E., Albarino, C. G., Khristova, M. L. & Nichol, S. T. Crimean-Congo hemorrhagic fever virus-encoded ovarian tumor protease activity is dispensable for virus RNA polymerase function. J Virol 84, 216–226, https://doi.org/10.1128/JVI.01859-09 (2010).
Beerens, N. et al. De novo initiation of RNA synthesis by the arterivirus RNA-dependent RNA polymerase. Journal of virology 81, 8384–8395 (2007).
Boonrod, K., Chotewutmontri, S., Galetzka, D. & Krczal, G. Analysis of tombusvirus revertants to identify essential amino acid residues within RNA-dependent RNA polymerase motifs. Journal of general virology 86, 823–826 (2005).
Pasternak, A. O., Spaan, W. J. & Snijder, E. J. Nidovirus transcription: how to make sense…? Journal of general virology 87, 1403–1421 (2006).
Stahl, M., Guba, W. & Kansy, M. Integrating molecular design resources within modern drug discovery research: the Roche experience. Drug discovery today 11, 326–333 (2006).
We would especially like to thank Prof. Outi Salo-Ahen, Åbo Akademi University, Finland for her useful suggestions during the project and for carrying out the QikProp property prediction (QikProp v5.8, release 12). Also, Prof. Mark Johnson and Biocenter Finland (bioinformatics) are acknowledged for the excellent computing facilities at the Åbo Akademi University. This study was supported by IRO scholarship Ph.D. Grant.