An in silico approach to analyze HCV genotype-specific binding-site variation and its effect on drug–protein interaction

Genotype variation in viruses can affect the response of antiviral treatment. Several studies have established approaches to determine genotype-specific variations; however, analyses to determine the effect of these variations on drug–protein interactions remain unraveled. We present an in-silico approach to explore genotype-specific variations and their effect on drug–protein interaction. We have used HCV NS3 helicase and fluoroquinolones as a model for drug–protein interaction and have investigated the effect of amino acid variations in HCV NS3 of genotype 1a, 1b, 2b and 3a on NS3-fluoroquinolone interaction. We retrieved 687, 667, 101 and 248 nucleotide sequences of HCV NS3 genotypes 1a, 1b, 2b, and 3a, respectively, and translated these into amino acid sequences and used for genotype variation analysis, and also to construct 3D protein models for 2b and 3a genotypes. For 1a and 1b, crystal structures were used. Drug–protein interactions were determined using molecular docking analyses. Our results revealed that individual genotype-specific HCV NS3 showed substantial sequence heterogeneity that resulted in variations in docking interactions. We believe that our approach can be extrapolated to include other viruses to study the clinical significance of genotype-specific variations in drug–protein interactions.


Results
Inter-genotype sequence comparison. To explore the inter-genotype sequence similarity in NS3 protein, multiple sequence alignment was performed using NS3 sequences from HCV genotypes 1a, 1b, 2b, and 3a, followed by the construction of sequence identity matrix. The analysis was carried out in two different ways, one using the full-length protein sequence (Table 1a) and the other employing only the fluoroquinolone binding region sequence (Table 1b). Both the approaches revealed an overall high degree of sequence similarity between NS3 helicases from genotypes 1a and 1b, i.e. 93.4% in the full-length protein sequence (Table 1a) and 94% in the fluoroquinolone binding region site (Table 1b). When compared to genotypes 2b and 3a, genotypes 1a and 1b revealed sequence identity scores ranging from 81.5-82.9% for full length protein sequence (Table 1a) and 83.3-88.1% for fluoroquinolone binding region site (Table 1b). Between genotypes 2b and 3a, sequence identity of 80.8% and 85.3% were observed, respectively, for full length and fluoroquinolone binding region site sequences (Table 1a and b).
Before inter-genotype mutation analysis (Fig. 1A), the genotype-specific consensus sequences were compared with the available reference sequences. The analysis revealed that the consensus and reference sequences were identical, with exception of 1-2 sites (Fig. 1B), validating that the consensus sequences used in our study represent 'true' genotype-specific sequences.
Our analysis revealed a high degree of conservation in helicase signature motifs across the four genotypes, except for motif V and motif VI (Fig. 1A), wherein certain genotypes a single amino acid substitution was observed (Fig. 1A). In motif V, F418Y variation was noted, where Y variant was found in genotype 2b; similarly, in motif VI, T459S was observed in genotype 1b, 2b, 3a, while K469R was observed in genotype 1b, 2b, 3a (Figs. 1A and 2). As mentioned in Methods, for this and other analyses 1A1V was used as the reference sequence.
Validation of 3D structures and docking strategy and genotype-specific variations in NS3-fluoroquinolone interactions. Before docking the NS3 structures were verified and validated using Verify 3D, GROMACS, and Ramachandran plot analysis. Structures constructed from three different tools were comparable, where structures constructed using Phyre 2 and Swiss-Model gave the lowest RMSD values of 0.41 and 0.0, respectively for genotype 2b and 0.42, and 0.07 for genotype 3a, respectively, when compared to the template structure. CPH tool has higher RMSD values of 1.55 and 1.53 for genotypes 2b and 3a, respectively, when compared to the template structures. All structures passed the 3D verification (performed using Verify 3D software) as at least 80% of the amino acids have scored > = 0.2 in the 3D/1D profile. Similarly, the structures were valid on the Ramachandran plot as most of the amino acids were under the permissible regions ( Supplementary Fig. 1). Table 1. Identity matrix of amino acid sequences of NS3 HCV of genotype 1a, 1b, 2b and 3a: Sequence identity was estimated in full length and fluoroquinolone binding region sequences by constructing an ID matrix using the 'BioEdit' software. The identity scores are expressed as percentages. www.nature.com/scientificreports/ Based on the lowest RMSD values, overall better quality (based on Verify 3D and Ramachandran plot assessment) models developed using Swiss-Model was used in further analyses (Fig. 3). The structural comparison revealed the structures of the four genotypes to be quite similar (Fig. 3). The RMSD values between structures of genotype 1a-1b, 1a-2b, 1a-3a, 1b-2b, 1b-3a, 2b-3a were found to be 0.01, 0.03, 0.03, 0.03, 0.03, and 0.04, respectively, suggesting strong homology between the NS3 structures from the four genotypes. Similarly, validation of the docking approach was done by performing blind docking (assuming drug binding site to be anywhere on the protein) on two previously reported complexes of NS3 bound to inhibitors, one a natural analog and other inhibitor ITMN-3479. Our results revealed that for both the models Molegro predicted the exact binding site/pose and gave the similar drug-protein interaction (for natural analog: Asp 454, Gln 481 and Cys 431; for ITMN-3479: Gly255, Thr269, Trp501) as reported for the reference models ( Fig. 4 A and B).
To investigate the effect of genotype-specific variations on NS3-fluoroquinolones interactions, molecular docking simulations were performed. Eight fluoroquinolones were docked individually on the consensus NS3 helicase structure constructed for each genotype.
Comparative analysis of the drug-protein interactions and binding poses revealed that almost all fluoroquinolones interacted with NS3 from all genotypes essentially in the same binding pocket, comprising amino acids Figure 2. Structure of HCV NS3 with genotype-specific amino acid substitutions across HCV variants 1a, 1b, 2b, and 3a. Ribbon Diagram of (A) front view, and (B) posterior view of NS3 reference structure (PDB ID: 1A1V). Within the fluoroquinolone binding region (P230-C584), amino acid residues variable across HCV genotypes are highlighted in yellow color, while the substrate-binding site, R393-W501, is shaded dark blue. Amino acids of genotype1a, 1b, 2b, and 3a are shown in the blue, red, green and purple font, respectively. A total of 73 amino acid variations were observed. To improve visualization, both the front and posterior views of NS3 were examined. Amino acids that were found to interact with fluoroquinolones are shown in orange boxes.

Discussion
In the current study, we present an in-silico approach to assess the effect of genotype-specific amino acid variations on protein folding and structural architecture and its interaction with the drugs. Our strategy involved computational analysis of NS3 helicase sequence and structures followed by molecular docking of fluoroquinolones on genotype-specific NS3 structures.
In the past era, a substantial number of studies explored the genetic heterogeneity of viral genotypes and its correlation with various aspects of the viral life cycle, including replication, and pathogenesis. Additionally, efforts have been made to unravel inter-genotype clinical and serological variations, functional differences in viral proteins and their effect on host immune system, generation of escape variants and viral epitopes, and on antiviral treatment response [16][17][18][19][20] . For example, the study conducted by Kaneez et al. showed amino acid variability in HCV NS3 1 and 3a and its impact on the structural architecture of the protein. The study also emphasized on the functionally active and conservative residues across selected HCV genotypes. To do so, several bioinformatics-based approaches and computational methods were employed. They reported that genotype-specific variations affected the structural architecture of NS3 21 . Similarly, Ahmed et al. 22 established the impact of sequence heterogeneity within NS5A and core regions on the response to pegylated interferon/ribavirin (PEG-IFN/RBV) therapy against HCV. The results showed a significant association between heterogeneity in IFN/RBV resistance-determining region IRRDR of NS5A and SVR, indicating that genetic heterogeneity in IRRDR can potentially serve as a predictor for SVR in HCV-infected patients treated with PEG-IFN/RBV combinatorial therapy 22 .
Another study published by Di Maio et al. implemented an in-silico approach to identify NS5B genetic variability in HCV genotypes and its potential effect on the genetic barrier for drug resistance (nucleoside inhibitors (NI)and non-nucleoside inhibitors(NNI)) 23 . The impact of these mutations was analyzed via docking of sofosbuvir with the NS5B protein of HCV genotypes. This study suggested that HCV sequence variability in NS5Bcan potentially alter the efficacy of NS5B inhibitors 23 .
The approach we have used here has the advantage that it analyzes the effect of naturally occurring genotypespecific amino acid variations on drug-protein interactions in the absence of drug-induced selection pressure. This method allowed us to correlate genotype-specific sequence heterogeneity with its influence on protein's structural architecture, and with its effect on variable interactions with selected drugs. This approach can be extended to other viruses to study the clinical significance of genotype-specific sequence variations that influence drug-protein interactions.
In the previous structural-functional studies of NS3 HCV, it has been proposed that the residues that play a critical role in NTP and substrate binding are found in the conserved helicase motifs: Motif I/Walker A (207′-GSGKSTK-213′), Ia (223′-YKVLVLNPSVA-233′), Walker B/Motif II (290′-DECH-293′), III (319′-VLA-TATPP-326′), IV (365′-LIFCHSKKK-373′), V (410′-ATDALMTGYTGDF-422′), and VI (456′-VSRSQRRG RTG R-467′)( Fig. 1) 21,24,25 . These helicase signature motifs are oriented in such a manner that their spatial arrangement forms the lining of NTP and substrate binding pocket, located in the cleft between domains 1 and 2 25,26 . Motif Y (Y241) is a characteristic motif of NS3 HCV located between motifs 1a and 1b. Tyrosine241 plays an important role in stacking the adenine base of ADP in correspondence with T419 26  In light of our observations, we speculate that A240S in genotype 3a, A431C in genotype 1a, T433E in genotype 3a, T435I/Y in genotype 2b and 3a, T445S in genotype 3a, E447T in genotype 2b, T449Q/R in genotype 2b and 3a, A500S in genotype 3a and F557L in genotype 1a might be of functional importance. These amino acids are present in the vicinity of the residues involved in the catalytic center and/or substrate binding loop, and therefore, can influence nucleic acid binding and NS3 catalytic activity. Residue Y241 is crucial for forming interactions with the adenine base of ADP 21,26 . In genotype 3a, we observed substitutions at both positions Scientific Reports | (2020) 10:20885 | https://doi.org/10.1038/s41598-020-77720-9 www.nature.com/scientificreports/ Balofloxacin ---ASN229   --PRO230  PRO230   --GLU291  GLU291   HIS293  HIS293  HIS293  HIS293   THR295  THR295  VAL295  GLN295   -ASP296  ASP296  ASP296   -SER297  ---THR298 -- Enrofloxacin Scientific Reports | (2020) 10:20885 | https://doi.org/10.1038/s41598-020-77720-9 www.nature.com/scientificreports/ Drugs 1a 1b 2b 3a Levofloxacin ---ASN229   ---SER231   --GLU291  GLU291   HIS293  HIS293  --THR295  THR295  VAL295  GLN295   ---ASP296   ---ALA297 HIS369 Scientific Reports | (2020) 10:20885 | https://doi.org/10.1038/s41598-020-77720-9 www.nature.com/scientificreports/ Drugs 1a 1b 2b 3a Sparfloxacin www.nature.com/scientificreports/ 240 and 241 simultaneously, while the Tyrosine241 was replaced with phenylalanine, alanine at position 240 that was substituted with serine. These substitutions concurrently can affect the binding of ATP molecule and subsequently the helicase activity due to changes in the electrostatic environment of the binding site. In close vicinity of Walker B Motif, lie residues 294 and 295. In this motif, H293 is of prime importance and is implicated in the coupling of ATPase and helicase activity 21 . Amino acid 294 and 295 are not only found in the proximity H293 but are also present at the entry side of the substrate. Therefore, substitutions at these positions, such as S294A and T295V/Q ( Figs. 1 and 2), might not only influence entry and binding orientation of ssRNA/DNA but may also affect NTP binding and hydrolysis. Motif V corresponds to the coordination of ATP and substrate binding 21,26,27 . We observed a substitution of amino acid F418Y, adjacent to the functionally important residue 419, involved in NTP binding and catalysis 21 . Here, we speculate little or no effect on the activity because the amino acid change was conservative. Residues 431C/C/A/A, 433T/T/T/E, 435T/T/I/Y, 445T/T/T/S, 447E/E/T/E, 449T/T/Q/R, 500A/A/A/S, 557L/F/F/F may play either a direct or indirect role in ssDNA/RNA interaction and unwinding. This leads us to speculate that genotype-specific variations in these amino acids might influence substrate interaction and helicase activity. A characteristic structural feature of NS3 HCV is a β-hairpin spanning the amino acids 430-452. This hairpin consists of two conserved phenylalanine residues F438 and F444. Residues V432 and T450 in the hairpin are known to be involved in nucleic acid interactions 28 . V432 intercalates between the DNA bases, stacks the nucleotide bases at 5′ terminal, and locks the enzyme in position, whereas T450 is involved in ssDNA binding to the nucleic acid pocket of the NS3 helicase 28 . F438 and F444 are responsible for the release ssDNA upon ATP binding 28 . It may be suggested that substitutions in the vicinity of these two residues, at positions 430T/T/V/V, 431C/C/A/A, 433T/T/T/E, 435T/T/I/Y, 445T/T/T/S, 447E/E/T/E, 449T/T/Q/R and 451L/V/V/A can alter the flexibility of the loop and subsequently might affect substrate interaction. Q460, a conserved residue of helicase Motif VI, plays a key role in ATP binding and coupling ATP hydrolysis with helicase translocation 26 . W501 is a critical residue for nucleic acid binding 26 .
To gain insights in drug-protein binding pattern across the HCV genotypes; molecular docking simulations were performed with NS3 protein models of each candidate genotype against a batch of 8 fluoroquinolones (Sparfloxacin, Ciprofloxacin, Balofloxacin, Levofloxacin, Lomefloxacin, Enrofloxacin, Pefloxacin, and Ofloxacin) previously shown to exhibit inhibitory potential against HCV NS3 helicase under in vitro conditions 12 . In our analysis, it was observed that the fluoroquinolones interacted with amino acids that constitute the helicase catalytic core via H-bond as well as non-bond interactions (Supplementary Table 2). Several studies have demonstrated that both H-bonds and hydrophobic interactions stabilize the ligands at the target site and assist in altering binding affinity and drug efficacy 29 . Furthermore, biological activity of the drug increased with increase in the number of hydrophobic interactions in the core of drug-target interface 29 . The residues that are functionally important and were found involved in H-bond as well as non-bond interactions with our drugs are P230, S231, T269, K371, R393, W501 and Y502 25,26 (Figs. 5 and 6; Table 2; Supplementary Table 2). Residues S231, T269, W501, and Y502 are functionally important as they are known to facilitate the interactions of NS3 with a 3′ segment of substrate/ssDNA. P230 and S231 are key residues involved in DNA binding. S231 interacts with substrate/ssDNA via bridging water molecule, while T269 is thought to be indispensable for helicase activity of NS3. K371 and R393, crucial residues of NS3 active site, are involved in substrate binding, where R393 makes interactions with the 5′ segment of ssDNA and coordinates in translocation of NS3 helicase 25,26 ; in addition to this, K371 participate in stabilizing the interaction with ssDNA/RNA 30 . W501 stacks the nucleotide bases near the 3′ end of the substrate/ssDNA and acts as a bookend along with V432 that defines a central binding cavity 26 .
Fluoroquinolones, Sparfloxacin (in genotype 3a), Balofloxacin (in genotype 2b and 3a), and Lomefloxacin (in genotype 3a) targeted residue P230 (Fig. 5 and Table 2) that might hinder in substrate binding. Enrofloxacin, Levofloxacin, and Pefloxacin (in genotype 3a) were found to interact with residue S231. Interactions of the Enrofloxacin, Levofloxacin, and Pefloxacin with these residues might affect ATP binding and transition as well as enzyme-substrate interaction. Likewise, Sparfloxacin (in genotype 1b) targeted T269 ( Table 2) that can potentially inhibit substrate binding and subsequently NS3 unwinding activity. R393 is thought to be a key residue that makes direct contact with ssDNA/RNA as well as facilitates nucleic acid unwinding 28 . Therefore, drugs such as Table 3. Molegro docking score for NS3-Fluoroquinolones interactions: Molegro docking score for top 10 poses observed for each fluoroquinolone-NS3 (specific to each genotype) interactions are shown. Previously determined in vitro IC50 (µM) for each fluoroquinolone, except for Levofloxacin (shown as -) is also given. www.nature.com/scientificreports/ Balofloxacin (in genotype 1a and 3a), Ciprofloxacin (in genotype 3a) targeting residue R393 (Fig. 5 and Table 2) can greatly affect substrate binding, NS3 translocation and subsequently nucleic acid unwinding. W501 is a critical amino acid of substrate binding cleft of NS3 HCV that serves as a bookend residue; stacking nucleotide of ssDNA/RNA at 3′ terminal preventing the protein from sliding along the nucleic acid and facilitates helicase translocation 26,28 . Ciprofloxacin, Pefloxacin and Enrofloxacin (in genotype 1b) and Levofloxacin and Lomefloxacin (in genotype 3a) interacted with W501, while most fluoroquinolones with interacted residues in their proximity of W501, namely E493 (present in genotypes 1b, 2b, and 3a) (Fig. 5 and Table 2), which might cause hindrance in nucleic acid interaction and can potentially inhibit the helicase activity of NS3. Maga et al. identified a series of compounds out of which QU663 exhibited a strong binding affinity with NS3. QU663 binds to the RNA binding cleft and makes direct contact with residues R393 and W501 31 that are crucial for substrate binding and helicase translocation. The present study showed that fluoroquinolones we tested also targeted the same crucial residues in the substrate-binding catalytic cleft, namely, W501 and R393 (Fig. 5).
We identify a few limitations of our study. A major limitation of this study was that our analysis was performed entirely in silico. We used two crystal structures (for 1a and 1b) and used a homology modeling approach to construct NS3 structures for genotypes 2b and 3a. Homology modeling heavily relies on the identification of the correct template, selected based on sequence identity 32 . A sequence identity above 50% generally tends to generate reliable structures with limited errors in loops and side-chain positing, while structures below 30% can have serious folding errors 32 . Additionally, regions in the query sequence that share low sequence similarity with the template protein, even though the rest of the protein show a high sequence similarity, can lead to erroneous folding. We tried reducing this bias/error by using three different homology modeling programs, validating and checking constructed structures for various errors and using a template that exhibited excellent similarity with our query sequences. Interestingly, all three programs identified 1A1V as a suitable template sharing more than 90% similarly with the sequence, and all three programs gave the same structures. Additionally, our structures passed the 3D verification (performed using Verify 3D software) as at least 80% of the amino acids have scored > = 0.2 in the 3D/1D profile. Similarly, the structures were valid on the Ramachandran plot as most of the amino acids were under the permissible regions. The other limitation we anticipate is with the molecular docking approach we used. In absence of experimentally characterized binding sites and experimentally derived binding energies, molecular docking analyses can be unreliable and can potentially miss out several key residues, present artificial interaction with residues and/or under-or over-estimate the binding scores 33 . We tried to address this issue by validating our docking approach using two previously reported structures bound to the ligand. We adopted a 'blind docking' approach, where the binding site of the ligand was not defined. Our software predicted the same pose for the ligand and also identified all the residues previously reported for the two structures. Another limitation that we identify is use of docking score to rank the poses and selection the top pose with highest docking score. Docking programs produce one (or several) different poses for every ligand, and further rank different compounds based on their scoring functions 34 . Several studies suggest that binding energies/scores predicted by the docking might be incorrect, despite the correctly predicted binding pose 34 . These can be overcome by using more robust approached such as MD simulations followed by WaterMap analysis; however, these approaches can be time-consuming and computationally demanding 34 . Nonetheless, as the experimentally solved structures continue to grow, the boundaries that differentiate between reliable and unreliable predictions will narrow and the capacity of the docking tools to predict correct poses/interactions will increase 32 . However, to gain further insights into the mechanism, site-directed mutagenesis of crucial residues followed by functional assays may be employed. These analyses should be supplemented by observations from the crystal structures of drug-enzyme complexes.
In conclusion, our approach provides an in-depth analysis of sequence variation in HCV NS3 protein across the selected genotypes and identifies active site amino acid residues crucial for drug-protein interactions. This approach can help study inter-genotype sequence heterogeneity in viral proteins and its correlation with the response to the antiviral treatment regimen.

Methods
Retrieval of HCV NS3 sequences and sequence analysis. A total of 1703 HCV NS3 sequences, belonging to genotype 1a, 1b, 2b, and 3a were downloaded in FASTA format from Los Alamos HCV Database (https ://hcv.lanl.gov/conte nt/index ) 35 . Other genotypes were not considered because their sequences were not present in the database in sufficient numbers to conduct a meaningful analysis. Out of these sequences, 687, 667, 101 and 248, belonged to, respectively, genotype 1a, 1b, 2b, and 3a. Sequences for each genotype were aligned separately and edited using the MEGA 6.0 software, implementing the ClustalW algorithm. The aligned nucleotide sequences were then used to generate consensus sequences for each of the candidate genotypes using the advanced 'Consensus maker' tool with 0.7 thresholds (i.e. residues represented in ≥ 70% of the sequences were used to generate a consensus) (http://www.hiv.lanl.gov/conte nt/seque nce/CONSE NSUS/conse nsus.html). In all genotypes, almost all consensus residues/sites were present in 99% of the sequences, with exception of a few sites, where consensus residues were present in 70-85% sequences (Supplementary Table 1). The consensus sequences of NS3 for genotypes 1a, 1b, 2b, and 3a were subsequently translated to amino acid sequences using the ExPASy translate tool 36 . Additionally, reference sequences for genotypes 1a, 1b, 2b, and 3a were downloaded in FASTA format PDB (genotype 1a: PDBid:1A1V and genotype 1b: PDBid:1CU1) and UniProt (genotype 2b: UniProt ID Q9DHD6 and genotype 3a: UniProt ID Q81495) databases. For genotype variation analysis, the genotype sequences were aligned with each other using the ClustalW algorithm implemented in MEGA 7 software 37 . Additionally, each genotype consensus sequence was also aligned, using MEGA 7, with its reference sequence to compare the difference in consensus sequences and the reference sequences. www.nature.com/scientificreports/ Inter-subtype sequence similarities and differences were determined by constructing a sequence identity matrix using 'Bioedit' software 38 . Two ID matrices were constructed: one with a full-length sequence of HCV NS3 helicase comprising 623 amino acids, and the other comprising of amino acids spanning Pro230-Cys584 (termed in this study as fluoroquinolone binding region). This region was selected because fluoroquinolones are found to form interactions with different amino acids within this region 39 . Additionally, this region also contains the previously reported helicase active site, spanning from Arginine393 to Tryptophan501 8,40,41 . Retrieval of HCV helicase structures and protein homology modeling. PDB and UniProt databases were searched for the available genotype-specific NS3 helicase structures and crystal structures were only found for genotypes 1a and 1b. These structures (genotype 1a: PDBid:1A1V and genotype 1b: PDBid:1CU1) were downloaded on .pdb format. For genotypes 2b and 3a, the Homology Modeling approach was adopted for the generation of 3D protein structures 10,42 . For accuracy and reliability, 3D models of NS3 were generated using three different homology modeling tools, namely the CPH model, the Swiss model and Phyre 2 [43][44][45] . The structures were visually inspected in Discovery Studio Visualizer version 4.0 (DSV4.0; Dassault Systèmes BIOVIA, Discovery Studio Visualizer, version 4.0, San Diego: Dassault Systèmes, 2016; Retrieved from http://accel rys.com/ produ cts/disco very-studi o/), and thereafter saved in PDB format. Subsequently, the structures were verified using the Verify 3D tool, while the energy minimization and validation were performed using the GROMACS tool 46 and Ramachandran plot analysis implemented in DSV4.0.
Fluoroquinolone structures. In this study, we used a panel of 8 fluoroquinolones, namely Sparfloxacin, Ofloxacin, Balofloxacin, Pefloxacin, Levofloxacin, Lomefloxacin, Ciprofloxacin, and Enrofloxacin, that have previously been reported to effectively inhibit HCV NS3 helicase 12 . Structures of these drugs were retrieved from PubChem Database in 3D SDF format 47 . Before docking analysis, SDF structures were converted to PDB format using DSV4.0.

Analysis of genotype-specific fluoroquinolone-NS3 interactions. Molecular docking simulations
were performed to analyze the genotype-specific protein-ligand interactions. The docking approach was validated by performing blind docking (assuming drug binding site to be anywhere on the protein) on two previously reported complexes of NS3 bound to inhibitors, one a natural analog M2 48 and other inhibitor ITMN-3479 (http://www.rcsb.org/pdb/explo re/litVi ew.do?struc tureI d=3RVB). Molecular docking studies and conformational analysis were conducted by Molegro Virtual Docker 49 and DSV4.0 software. Molegro Virtual Docker binding site was set to default, where the Molegro identifies the binding ligand binding sites in the protein. For our analysis, Molegro identified two binding sites, one located in the DNA binding domain, which contained the key residues within the active site of NS3, as identified by previous reports 8,40,41 , and others in the ATP domain of the helicase. The 8 fluoroquinolones were docked individually on each of the four genotype-specific models of NS3, into the identified cartesian space. Molecular docking was performed using standard precision protocols with default parameters of Molegro Virtual Docker. A total of 2000 interactions were simulated, and out of those top 10 poses was selected based on docking energies. Using these poses, analysis of genotype-specific variations on NS3-fluoroquinolone binding was carried out. Visualization of docking poses and analysis of drug-protein interactions were performed using DSV4.0.