In silico HPgV NS3/NS5B Models Show Predicted Binding of Several HCV Antivirals

Investigators reporting on human pegivirus (HPgV) used structure-based modeling to compare two essential replication enzymes—NS3/4A protease and NS5B RNA-dependent RNA polymerase—with their hepatitis C virus (HCV) counterparts, then ran a high-throughput docking screen of approved small molecules. Within that workflow, the authors highlighted several HCV direct-acting antivirals (DAAs) with favorable predicted binding in conserved catalytic pockets of the modeled HPgV proteins. They describe the combined modeling, simulation, and docking results as structural evidence meant to prioritize candidates for experimental follow-up, rather than to support clinical claims. Overall, the study outlines a computational prioritization strategy based on predicted structural and interaction similarity between HPgV and HCV targets.

To generate the HPgV structures used downstream, the team built NS3/4A and NS5B models using SWISS-MODEL homology modeling and produced independent structures with AlphaFold2 from the same consensus sequences, with HCV crystal structures serving as the experimental reference points for comparison. For the SWISS-MODEL outputs aligned to HCV experimental structures, the reported backbone RMSD values were 1.00 Å for NS3/4A and 1.26 Å for NS5B. The article also describes a template-selection step based on SWISS-MODEL’s Global Model Quality Estimate (GMQE), with templates below a prespecified threshold excluded before proceeding. These alignments and template filters were presented as the rationale for carrying the modeled pockets forward into simulation and docking analyses.

Model behavior over time was examined using 100 ns molecular dynamics simulations of the apo (ligand-free) HPgV NS3/4A and NS5B structures. As reported, the authors summarized trajectory stability using standard descriptors including RMSD, root-mean-square fluctuations (RMSF), and radius of gyration, alongside a qualitative description of regions characterized as more flexible (for example, termini and selected loop segments) versus more ordered. They also report comparatively stable catalytic-region geometry across the simulations, including the NS3 catalytic triad and the NS5B GDD motif region. In the paper’s framing, these dynamics results supported using the modeled active sites as docking targets within the study’s in silico workflow.

With those models in place, the authors screened a DrugBank-derived set of 2203 FDA-approved small molecules by docking into catalytic-site grids centered on conserved residues within HPgV NS3/4A and NS5B, using AutoDock Vina as the docking engine (as reported in this study). They then compared docking scores between HCV experimental structures and the HPgV models across the full library to quantify similarity in predicted affinity patterns. In that analysis, Pearson correlations were reported between HCV and HPgV docking affinities of r = 0.46 for NS3/4A and r = 0.76 for NS5B (across compounds). The article presents these correlations as a numerical summary of how similarly the two viruses’ targets behaved within the docking framework.

In the overall screen of FDA-approved small molecules, only one HCV DAA (paritaprevir) appeared among the top docking hits; other DAAs (e.g., glecaprevir, voxilaprevir, grazoprevir, dasabuvir, and sofosbuvir) were evaluated separately and showed favorable predicted affinities in the targeted analyses. For NS3/4A, the paper describes hydrogen-bonding and hydrophobic contacts involving residues near the catalytic pocket, with one example reporting glecaprevir predicted at −11.5 kcal/mol and forming hydrogen bonding with SER154 in the HPgV NS3 model. For NS5B, the reported interaction mapping emphasized contacts near conserved polymerase motifs, including the GDD motif region and a described hydrogen bond involving ASP308 for docked ligands such as dasabuvir and sofosbuvir. The authors present these residue-level interaction diagrams as hypotheses for which molecules might be prioritized for biochemical testing.

The discussion also notes caveats the authors associate with interpreting docking scores, including the potential for false-positive high-affinity predictions when ligands are forced into large, hydrophobic grooves and the broader limitation that in silico pocket scoring does not directly capture access, conformational states, or experimental binding behavior. They frame additional wet-lab work—such as biochemical or in vitro validation—as the appropriate next step to evaluate whether prioritized compounds inhibit HPgV replication targets, while noting that their current results are based on consensus-sequence models without experimental HPgV structures. The paper further describes the resulting models and docking outputs as possible inputs for computational candidate-prioritization workflows, including downstream triage approaches that could incorporate machine-learning methods. Taken together, the reported scope remains a computational screen built around modeled structural conservation and predicted active-site interactions.

Key Takeaways:

• These findings are based on in silico homology modeling, docking, and molecular dynamics simulations, without experimental HPgV structures or binding data.
• HPgV NS3/4A and NS5B models were generated with two structure-prediction approaches and were reported to align closely to HCV experimental structures, which the authors used to justify downstream simulations.
• A DrugBank-derived library of FDA-approved small molecules was docked into modeled HPgV catalytic pockets; docking-score patterns were reported to correlate between HCV and HPgV targets, and multiple HCV DAAs were reported to show favorable predicted binding near conserved catalytic residues/motifs as candidates for experimental follow-up rather than clinical recommendations.