1 Introduction

The viruses of the family Coronaviridae are now notoriously famous for their diseases causing capabilities in birds, humans and mammals. The corona virion typically composed of RNA enclosed in enveloped protein, having glycoprotein spikes, is capable of infecting a broad range of hosts, including humans. Coronaviruses, as the number of variants and diversity increases in this family, based on similarities are classified into four sub-genera, designated as alpha (α), beta (β), gamma (γ) & delta (δ) [1]. So far, the β coronaviruses (CVs) are known to cause infections in humans, including common colds and primarily affecting the respiratory system. Bats are associated with the CVs pandemics in the human population, bats harbor the virus and are believed to be immune to the viral infection itself, promoting the mutations that are crucial for the CVs pathogenicity [2]. The spike-like glycoprotein (S), giving the virus its corona like appearance is vital for their pathogenicity and helps them to attach with the host cell surface receptors and also delimits the hosts’ range for the CVs [3].

The CVs genome, ranging from 27 to 32 kilo-bases, are positive-sense single-stranded RNA (+ ssRNA) coding for, ORF1a and ORF1b, the poly-proteins involved in RNA polymerization (RNA-dependent RNA-polymerases) (RdRp) and also for modulation of host responses [4, 5]. Fatal diseases causing zoonotic strains in this family are severe acute respiratory syndrome (SARS) and the Middle East respiratory syndrome (MERS) [6]. Additionally, there are four more strains, which are reported to be disease-causing in humans, mainly common colds in individuals with immunodeficiency (229E, HKU1, NL63 and OC43) [4].

The 2019-novel-corona-virus (SARS-CoV-2) that emerged in Wuhan in 2019 belongs to a bat derived Coronaviridae family, that have gained the transmission capability from animals to humans and from human to human, due to which SARS-CoV-2 became so lethal and caused global emergency [7]. The SARS-CoV-2 is an enveloped RNA virus with the distinctive corona like shape protein spikes (usually about nine to twelve nanometers) capable of attachment to host cells. The SARS-CoV-2 potentially causes “novel corona-virus-infected pneumonia” or NCIP, the disease of lower respiratory tract having common cold-like symptoms with fever chest congestion leading to difficulty in breathing [8].

The SARS-CoV-2 has an 86.9% similarity with the genome of bat-like SARS-CVs and was classified as a distinctive subclade in the subgenus of sarbecovirus having typical β-CVs genome organization [8]. The SARS-CoV-2 genome, like other CoVs constitutes a 5ʹ untranslated region (UTR) replicase-complex Orf1a and Orf1ab followed by protein-encoding genes for the spike (S), membrane (M), envelope (E), nucleic capsid (N) and a 3ʹ UTR [9].

The non-structural proteins (nsp) from 1 to 16 of CoVs have a vital role in their replication, while the functions of certain nsps remain elusive. The structural proteins are indispensable for viral assembly and infection, while S protein for spike has distinctive variations and helps in the attachment to the host cell surface proteins [10, 11]. The M protein having transmembrane domains binds to the nucleocapsid and shaping the virion [12, 13]. The E protein is indispensable for viral pathogenesis and is responsible for virion assembly and budding [14, 15]. The N protein is comprising of two domains, having the capability of binding with virion genome and nsp-3 protein triggering replicase-transcriptase complex and viral genome encapsulation [16,17,18].

Herein, we used a multi-steps computational pipeline to identify novel compounds against the RdRp from SARs-CoV-2. Virtual screening, docking and re-docking approaches were used followed by molecular dynamics simulation and free energy calculation. Novel hits were identified, which possess better inhibitory properties than Remdesivir. This disclosing of necessary knowledge regarding the screening of natural products and the identification of top hits could help to design effective therapeutic candidates against the coronaviruses and design robust preventive measurements.

2 Material and Methods

2.1 Phylogenetic Analysis of Coronavirus:

NCBI database was used for the retrieval of Corona viruses (RdRp region) sequences. The accession no MT042778 was used as query sequence in the NCBI Blast for obtaining highly similar sequence. For the selected sequences either we can download complete sequences or only download aligned sequences using options available in NCBI. In this study we downloaded only aligned sequence to make sure to get only RdRp region. Total 110 sequences were retrieved, 107 sequences of SARS CoV-2 were placed as ingroup and remaining three of sequences Bat-CoV were used as outgroup. These sequences were aligned with the help of Clustal software [29] using pairwise multiple sequence alignment algorithm. This data matrix was used for generating trees file using Beast software [30]. Three independent Markov chain Monte Carlo analyses of 100,000,000 steps were conducted and one best tree was saved after 2000 steps. The effective sample size (ESS) of all parameters above 200 is an indication of reliable results, which is checked on Tracer ver 1.5 [31]. Trees file was uploaded to tree annotator software for obtaining a Maximum clade credibility tree with posterior probability and branch length information. The annotated tree was visualized on Figtree software [32].

2.2 Protein Structure Preparation and Active Site Identification

A cryo-EM structure of RdRp from SARs-CoV-2 was downloaded from RCSB using PDB ID: 6M71 [19]. The structure was subjected to energy minimization and missing residues by using the protein preparation wizard implemented in Schrodingers Maestro [20]. Structural topology was reviewed for the defects. MolProbity [21] was used to assess the quality of the constructed structure, and energy minimization was used to resolve atomic conflicts using steepest descent and gradient conjugation algorithms. The water molecules were stripped and visualized in PyMOL [22]. The active site of the RdRp is located in the seven conserved motifs from A to G. SDD sequence (residues 759, 760 and 761) K545 and R555 are reported to be a potential target for drug discovery [23]. Thus based on these residues, the active site was selected and used for virtual screening and molecular docking.

2.3 Ligands Database Retrieval, Preparation, and Virtual Screening Protocol

The database of compounds from medicinal plants from Northern south Africa was retrieved from NANPDB (https://african-compounds.org/nanpdb/) [24]. This database is a diverse source of natural drugs from 617 source species, which comes from 146 families of plants, animals, bacteria, and fungi. A SDF format file was downloaded, which comprised of 6482 compounds. Structural preparation such as charges, minimization, and compound washing was carried out. The database was then converted to.mdb format to be used as input for Molecular Operating Environment (MOE v2016) [25]. The selected residues option was used to define the active site residues. With ten conformations, each ligand was screened against the active site using a triangle matcher as a placement while London dG as a scoring method. Docking scores and visual interactions were used as a criterion for selecting the best hits.

2.4 Molecular Docking and Re-docking

Compounds obtained from virtual screening were subjected to further screening for the best active compounds against the RdRp active site. Prior to molecular docking, the obtained hits were subjected to pharmacokinetics and pharmacodynamics criteria validation, which excluded 36 compounds. The obtained 199 compounds were subjected to induced-fit docking protocol using MOE. Using the IFD protocol, the compounds were further reduced to a reasonable number, which could be then evaluated and docked individually against the active site of RdRp. For the re-docking, we used AutoDock Vina software, which is based on a Genetic Algorithm (GA) [26]. AutoDock software was used to define the grid dimension and box based on the defined residues. The protein structure was converted to .pdbqt format while using the ligands preparation criteria such as root detection, charges, hydrogen, and aromaticity criteria were used for ligands preparation. Each ligand molecule was prepared individually and converted to. pdbqt file. To achieve high accuracy, we set exhaustiveness to 64. To compare our docking results, we used Remdesivir as control. Thus here, a multi-steps docking and re-docking approaches were utilized to identify the most potential hits that could probably bypass Remdesivir in both computational and experimental setups to inhibit the SARs-CoV-2. To predict the bioactivity of each top ligand Molinspiration Cheminformatics tool was used while for ADMET analysis, SwissADME was utilized [27].

2.5 Simulation Protocol

The identified top hits and Remdesivir complexes were subjected to molecular dynamics simulation to understand the dynamics and interacting behavior of these compounds. To obtain better and accurate simulation results, all the complexes were submitted to Propka 3.1, which is an online web server, for the correction of the protonation state. Amber 18 package with pmemd.cuda implementation was used to perform the simulations [28]. The latest AMBER ff14SB force field was used for simulation. The antechamber was used to prepare the ligand topologies and obtained.frcmod file for simulation [29]. The generalized Amber force field (GAFF2) was used for small molecules parameters, and Gasteiger charges were added to each inhibitor [30].

TIP3P water box with 14 Å buffer distance each side was used to solvate the systems. Each system was neutralized by adding Na + ions. Using the 300 K temperature controlled by Langevin thermostat and a pressure of 1.0 bar scrutinized by Berendsen Barostat was used for each system [31, 32]. All bond lengths involving hydrogen atoms were constrained by the SHAKE algorithm [33]. A time step was set as 2.0 fs. For long-range interactions, particle mesh Ewald summation (PME) approach was exercised [34]. For all cases, the non-bonded cut-off was fixed at 10.0 Å. Each system was minimized by using two-step minimization approach. Followed by heating and equilibration, the production simulation was carried for 100 ns at the NPT ensemble, and the Cartesian coordinates were stored at every 10 ps. Overall, 10,000 frames were obtained from each production simulation.

2.6 Post-Simulation Analysis and Visualization

The trajectories obtained from each system were subjected to post-simulation analysis such as root mean square deviation (RMSD) to estimate the stability of each system, root mean square fluctuation (RMSF) to access the flexibility at residues level. For structure compactness, we calculated the radius of gyration as criteria for determining the structural compactness during the simulation time. For all these analyses, we used CPPTRAJ and PTRAJ [35].

2.7 Binding Free Energy Calculations

The binding of each ligand was estimated by using the molecular mechanics Poisson–Boltzmann surface area (MMGBSA) method which is a widely used and acceptable method [36,37,38,39,40]. The most widely used MMPBSA.py script was used as input, which contain all the guidelines for free energy calculations. For each system, 2500 structural frames were used to calculate the free energy using the following equation.

$$\Delta G_{{{\text{bind}}}} = \, \Delta G_{{{\text{complex}}}} - \, \left[ {\Delta G_{{{\text{receptor}}}} + \, \Delta G_{{{\text{ligand}}}} } \right]$$

In this equation, ΔGbind represents total free binding energy, while others show the free energy of complex, the protein, and the ligand. Specific energy term contributes to the whole free energy was calculated by the equation:

$$G = G_{{{\text{bond}}}} + \, G_{{{\text{ele}}}} + \, G_{{{\text{vdW}}}} + \, G_{{{\text{pol}}}} + \, G_{{{\text{npol}}}} - TS$$

Gbond, Gele and GvdW specify interactions among bonded, electrostatic, and van der Waals states. In contrast, Gpol and Gnpol represent the polar and non-polar interaction to the free energy presumed through precise GB (Generalized Born). This free energy calculation method is widely used by different studies to understand the binding energy of different ligands [41, 42].

3 Results

3.1 Phylogenetic Analysis

Phylogenetic analysis was conducted using molecular data (RdRp region) for the estimation of evolutionary relationship among SARS-CoV-2 members sampled from various geographical regions. In clade I the early branches are occupied by the members of USA, China and Thailand, which possibly suggest the evolution of SARS-CoV-2 in these areas. The oldest branches in clade II are occupied by the members having distribution in Thailand, which suggests that the ancestors of clade II evolved in Thailand. The members from Spain (ESP) and Jamica (JAM) occupied the early branch in clade III and IV respectively. Plesiomorphic (primitive) branches in clade V are occupied by the members of USA and Italy. This study indicates that there are two possible centers which are important in the origin and dispersal of SARs-CoV-2, these centers are East Asian center (China + Thailand) and North American center (USA). A secondary center Italy + Spain was also instrumental in proliferation of this virus.

The members of SARS-CoV-2 from Pakistan, Iran and Israil are grouped with the members from USA, whereas the Indian members are in group with Spanish and European members. Most of East Asian members are nested within Chinese members. Australian members are on same branches with Indian, East Asian and USA members (Fig. 1).

Fig. 1
figure 1

Phylogenetic tree constructed by Beast. The values above nodes are posterior probability values. Clade I–V are discussed in this study

Full size image

3.2 Virtual Screening and Molecular Docking

A multi-steps drug screening approach was used to search for the most potential drug candidate against the RdRp from SARs-CoV-2. A total of 6842 drugs from South African natural resources were screened in three steps. In the first step using MOE, all the compounds were screened, and the scores obtained from this screening range from − 7.0 to − 3.0 kcal/mol. To select the best compounds from all these, a criterion based on docking score and multiple interactions with the defined active site residues was used to filter the top hits. This screening resulted in 236 best compounds satisfying the specified criteria. Each conformation was manually visualized for this purpose. The obtained 236 compounds were then subjected to ADMET analysis, which excluded 37 compounds while the remaining were the best fit. Using the IFD methods, the remaining 199 compounds were again screened against the RdRp polymerase. In the case of Induced fit docking, the scores obtained were range from − 8.16 to − 4.34 kcal/mol. Here we again follow the same criteria to select the best hits using molecular docking score and visual interaction analysis. From these, 199 compounds, only 42 compounds were found to form the best interactions with active site residues and to have good binding affinity.

To further validate the activity of these final hits against the RdRp, we used the best algorithm (Genetic Algorithm) by AutoDock Vina. These 42 compounds and the receptor molecules were prepared and converted to the AutoDock Vina acceptable format (.pdbqt). Exhaustiveness was set 64 to achieve high accuracy. Results from AutoDock Vina range from − 10.4 to − 5.1 kcal/mol (Table 1). Considering Remdesivir as control, the docking score reported by AutoDock Vina was − 7.1 kcal/mol. Hence, using − 7.1 kcal/mol as a threshold, 24 compounds were found to have a docking score better than the Remdesivir docking score. Finally, these 24 compounds were analyzed, and the top 4 hits with the best docking score and Remdesivir were selected for further comparatively analysis.

Table 1 Docking of the top 42 compounds using AutoDock Vina
Full size table

3.3 Interaction Analysis of Top Hit and Remdesivir with RdRp

Analysis of the top hits and Remdesivir revealed that all the compounds possess strong inhibitory effects against the RdRp. In the case of Remdesivir, the docking score was found to be − 7.1 kcal/mol. As given in Table 2, Remdesivir forms five hydrogen bonds with the key active site residues. These residues include Lys621, Cys622, Asp761, Lys798, and Glu811. Besides, five hydrophobic and one salt bridge was also formed with different residues. On the other hand, the best compound Diosmetin-7-O-Beta-d-apiofuranoside with docking score − 10.4 kcal/mol formed nine hydrogen bonds with the key active site residues including Trp617, Tyr619, Lys621, Cys622, Asp623, Asp760, Asp761, Ala762, and Trp800. Furthermore, hydrophobic interactions with the key residues Asp618, Lys798, and Glu811 were also observed. Among the top four hits identified, the second compound 3-O-Alpha-l-arabinopyranosyl-echinocystic acid formed eight hydrogen bonds with Asp452, Thr556, Asp618, Tyr619, Lys621, Asp623, Arg624, and Asp760. Alongside salt bridges and π-Cation interactions with Arg553, Lys621, and Lys798 were also formed. The docking score for 3-O-Alpha-l-arabinopyranosyl-echinocystic acid was reported to be − 9.9 kcal/mol. Furthermore, compound 3′-epi-afroside with the docking score − 9.3 kal/mol also formed eight hydrogen bonds with Trp617, Tyr619, Lys621, Cys622, Asp623, Asp760, Asp761, Trp800 and three hydrophobic interactions with Asp618, Lys798 and Glu811 were observed. However, no salt bridge or π-Cation interaction was reported. Among the top-scoring best hits, Genkwanin 8-C-beta-glucopyranoside was also included. The docking score for the 4th ranked compound was reported to be -9.1 kcal/mol. Seven hydrogen bonds with the key active site residues such as Asp452, Arg553, Thr556, Lys621, Cys622, Asp623, Asp760, and one salt bridge with Arg555 was observed. These results are self-explanatory that the compounds identified through a multi-step screening and docking possess better inhibitory effects than Remdesivir. Not only these compounds possess the best docking scores, but also multiple interactions with the key amino acids are observed. The interaction pattern of all these compounds, including Remdesivir used as a control, are given in Fig. 2. Furthermore, details including the drug names, final docking score, interactions which includes hydrogen bonding, hydrophobic interaction, salt bridges, and π-Cation interactions are given in Table 2.

Table 2 The table is showing the results obtained from virtual screening and a controlled drug
Full size table
Fig. 2
figure 2

Interaction pattern of RdRp from SARs-CoV-2 with Remdesivir and the top four hits from the Northern African Natural products database. a Remdesivir, b Diosmetin-7-O-Beta-d-apiofuranoside, c 3-O-Alpha-l-arabinopyranosyl-echinocystic acid, d 3′-epi-afroside and e Genkwanin 8-C-beta-glucopyranoside

Full size image

Furthermore, the bioactivity of these top hits and Remdesivir was predicted and compared. As given in Table 2, the bioactivity predicted by the Molinspiration Cheminformatics tool reported that the top four hits possess strong bioactivity against enzymes than Remdesivir. The bioactivity score for Remdesivir was reported to be 0.38, which is the same as 3-O-Alpha-l-arabinopyranosyl-echinocystic acid (0.38). The compound Diosmetin-7-O-Beta-d-apiofuranoside was reported to have the bioactivity score 0.36 while 3′-epi-afroside and Genkwanin 8-C-beta-glucopyranoside possess many fold stronger bioactivity than Remdesivir. The bioactivity scores for these two compounds were reported to be 0.75 and 0.40, respectively.

Furthermore, the ADMET properties of such as molecular weight, LogP, number of rotatable bonds, hydrogen bond donor, and acceptors were calculated for each compound. It can be seen all the four compounds obey the ADMET properties and thus increases the reliability of experimental results. All the results are given in Table 3.

Table 3 2D structures, ADMET properties, and bioactivity of the top 4 hits and Remdesivir. The Molinspiration server predicts the activity of the compounds against different classes. If the score is between 0 and 5, it is considered as the best
Full size table

3.4 Dynamics Stability and Flexibility Analysis

To understand the dynamics stability and convergence, RMSD as a function of time of all the systems was calculated. The RMSD of all five systems is given in Fig. 3. It can be scene that the Diosmetin-7-O-Beta-d-apiofuranoside complex reached a stable equilibrium after 20 ns. The system possesses stable behavior during simulation. The average RMSD for the Diosmetin-7-O-Beta-d-apiofuranoside system was observed to be 2.0 Å. On the other hand, a little convergence between 20 and 50 ns was observed in the case of 3-O-Alpha-l-arabinopyranosyl-echinocystic acid, but soon after 50 ns, the RMSD values fell, and the average RMSD was observed to be between 2.5 and 3.0 Å. Comparatively, this system remained relatively unstable than the first one. Likewise, the two other systems also remained stable during the simulation. In the case of 3′-epi-afroside, a little convergence between 60 and 70 ns was observed, but overall the system remained stable. The average RMSD for 3′-epi-afroside was decreased to be between 2.0 and 2.5 Å. However, in the case of Genkwanin 8-C-beta-glucopyranoside, the systems showed acceptable convergence at different intervals, but overall the system was stable. For Genkwanin 8-C-beta-glucopyranoside, the average RMSD was to be between 2.0 and 2.5 Å. We also simulated the Remdesivir complex to understand its behavior. In the case of Remdesivir, the average RMSD remained higher than the other. At different time intervals, acceptable convergences were observed too. Overall, these results suggest that the identified compounds possess stable behavior during the simulation.

Fig. 3
figure 3

RMSD of all the five systems. a Remdesivir, b Diosmetin-7-O-Beta-d-apiofuranoside, c 3-O-Alpha-l-arabinopyranosyl-echinocystic acid, d 3′-epi-afroside and e Genkwanin 8-C-beta-glucopyranoside. The x-axis shows time in nanosecond while y-axis shows RMSD in Å

Full size image

Furthermore, we also determined residual flexibility by using Root Mean Square Fluctuation (RMSF). It is evident from Fig. 4 that all five systems display more or less similar fluctuations. In the case of Diosmetin-7-O-Beta-d-apiofuranoside and 3′-epi-afroside systems, a higher fluctuation between 50 and 80 residues can be seen while no significant differences in other regions are observed. On the other hand, Remdesivir and Genkwanin 8-C-beta-glucopyranoside posses similar fluctuation with increased fluctuation between 100–180 and 450–500 residues. In the case of Genkwanin 8-C-beta-glucopyranoside, a little higher fluctuation between 10 and 30 amino acids was observed. Furthermore, 3-O-Alpha-l-arabinopyranosyl-echinocystic acid showed a different fluctuation between 160 and 180, which is very different from others. Thus, the binding of these ligands differentially affects the internal dynamics and residual flexibility.

Fig. 4
figure 4

RMSF of all the five systems. a Remdesivir, b Diosmetin-7-O-Beta-d-apiofuranoside, c 3-O-Alpha-l-arabinopyranosyl-echinocystic acid, d 3′-epi-afroside and e Genkwanin 8-C-beta-glucopyranoside. The x-axis shows the total number of residues while the y-axis shows RMSF in Å

Full size image

3.5 Radius of Gyration (Rg) Calculation

The structural compactness of each system was analyzed by estimating the radius of gyration (Rg) from their respective MD trajectories, and the average values are reported. A similar Rg is obtained for all the four systems except the Remdesivir System. The average Rg for top hits systems was found to be between 25.2 and 25.4 Å, while this value increased for Remdesivir, and the average value was reported to be 25.8 Å. Thus the four compounds (top hits) bound to RdRp imply sustained stability and compactness of the complexes. Alternatively, the higher Rg value in the case of Remdesivir than the others, causing the interactions between ligand and protein to be weaker. Thus, we speculate that these compounds explored through computational pipeline may possess robust inhibitory effects than Remdesivir in the experimental assays. All the Rg(s) calculated are given in Fig. 5.

Fig. 5
figure 5

Rg of all the five systems. a Remdesivir, b Diosmetin-7-O-Beta-d-apiofuranoside, c 3-O-Alpha-l-arabinopyranosyl-echinocystic acid, d 3′-epi-afroside and e Genkwanin 8-C-beta-glucopyranoside. The x-axis shows the total number of frames while the y-axis shows Rg in Å

Full size image

3.6 Binding Free Energy

To estimate the binding free energy of each complex, a MM/GBSA was used. MM/GBSA is the most popular and reliable approach to calculate the binding energy of ligand during MD simulation. The total binding energy of a system calculates different energy terms such as SASA, vdW, PS, and electrostatic energy. To compare the results of the top four hits, Remdesivir, which is considered as the most potent drugs reported being active against the RdRp. The results confirmed that the four hits identified from screening, docking, and re-docking possess better binding affinities than Remdesivir. It was reported that Remdesivir possesses the total binding energy − 54.4061 kcal/mol. Whereas the other four possess − 59.486 kcal/mol (Diosmetin-7-O-Beta-d-apiofuranoside), − 57.184 kcal/mol (3-O-Alpha-l-arabinopyranosyl-echinocystic acid), − 60.315 kcal/mol (3′-epi-afroside) and − 65.695 kcal/mol (Genkwanin 8-C-beta-glucopyranoside) respectively. While the other energy terms such as van der Waals energy, electrostatic energy, polar solvation energy, solvent‐accessible surface area are given in Table 4, thus these results strongly suggest that the top hits identified here should be tested experimentally against the SARS-COV-2 at earliest.

Table 4 Shows the total binding free energy and related term of all the five complexes subjected to MMGBSA analysis
Full size table

4 Discussion

RNA-dependent RNA-polymerase is an important replicating enzyme which plays important role in the processing of RNA from SARs-CoV-2. The cry-EM structure of the RdRp recently reported revealed that the structure possesses similar architecture of Finger, Palm, Thumb and NiRAN region. A higher identity between the previously reported SAR-CoV and the recently reported structure is due to high amino acid conservancy. The study highlighted important residues, domains, and conserved motifs will help to identify potent inhibitors and help to control the emerging infections related to Coronaviridae family [19].

Computational methods are of great importance in determining the structure and function of proteins, drug binding, exploring the resistance mechanism, and bio-catalysis [41,42,43]. So, herein, using structure-based virtual screening approach shortlisted the top hits which forms important hydrogen, hydrophobic and other important interactions with the RdRp active site residues. The top hits were confirmed by performing IFD, which further shortlisted the top hits list very precisely. Using another round of docking with different algorithm exempted further hits from the list and shortlisted the top hits which could bypass Remdesivir. The use of molecular dynamics simulation technique and free energy calculations is the most widely practiced approaches while studying the protein ligand interaction. Integrating this pipeline further increased the reliability the quest to test our top hits experimentally because of its promising results. Thus, this study comprised of a complicated and multiple validations stress on the experimental assays of the top hits to help to contain the recent outbreak.

5 Conclusion

In conclusion, this study identified novel hits from natural sources. Using the structure-based approaches shortlisted the top hits which could inhibit this target experimentally. Furthermore, we also validated our shortlisted compounds by using simulation and free energy calculation. Thus, this study is a significant consideration in future strategies against the outbreaks caused by such viruses.