MOLECULAR MODELING AND VIRTUAL SCREENING FOR COMPUTER-AIDED DESIGN OF HCV NS5B POLYMERASE INHIBITORS

Bertrand-Ulrich Yavo1image, Akori Elvice Esmel1image, Kouakou Jean-Louis Kouakou1image,

Melalie Kéita1image, Megnassan Eugène1,2*image

¹Laboratory of Fundamental and Applied Physics, UFR SFA, Nangui Abrogoua University, BP 801 Abidjan 02, Côte d’Ivoire.

2ICTP-UNESCO, QLS, Strada Costiera, 11, I-34151, Trieste, Italy.

 

Abstract

Aims and objectives: This study aims to design new inhibitors for the Hepatitis C Virus (HCV) Non-Structural Protein 5B (NS5B) polymerase, an enzyme essential for viral replication. The research addresses an urgent public health issue affecting 71 million people and causing approximately 242,000 deaths annually.

Methodology: The research follows a computer-aided rational design approach. A Quantitative Structure-Activity Relationship (QSAR) model was developed using 24 quinazolinone derivatives (QDs) to correlate Gibbs free energy with experimental inhibition constants. The bound conformations of the ligands were used to construct a 3D-QSAR pharmacophore (PH4) model. A virtual library of 168,750 QDs was generated and filtered using ADME (Absorption, Distribution, Metabolism, and Excretion) criteria and PH4 screening. Conformational stability was evaluated through 200-ns molecular dynamics (MD) simulations. Binding free energy variations were quantified using the Molecular Mechanics - Generalized Born Surface Area (MM-GBSA) approach on MD trajectories, calculating molecular mechanics energy, solvation energy, and surface area contributions under the OPLS2005 force field.

Results: The QSAR model showed high predictive power and the PH4 model achieved an R2 of 0.85. Screening identified 39 potent analogues. The lead candidate, 3-6-4-45, exhibited a predicted inhibitory concentration of 0.62 nM, approximately 96 times more active than the best reference ligand (60 nM). MD simulations confirmed stability with RMSD values between 1.5 and 3 Å. MM-GBSA binding energies converged with predicted complexation energies, validating the computational reliability.

Conclusion: The integration of molecular modeling and in silico screening successfully identified six potent candidate inhibitors of the HCV NS5B polymerase with favorable pharmacokinetic profiles. These analogues represent high-affinity candidates for future therapeutic development.

Keywords: Computer-aided rational design, inhibitors, molecular dynamics, NS5B polymerase, pharmacophore, QSAR, quinazolinone derivatives, virtual screening. 

 

 

INTRODUCTION

 

The liver is an essential organ that performs vital physiological functions. Its impairment by chronic viral infections, such as the hepatitis C virus (HCV), leads to severe conditions, including cirrhosis and hepato-cellular carcinoma1. HCV remains a significant global public health issue, affecting approximately 71 million individuals and resulting in hundreds of thousands of deaths annually due to liver-related complications2. Despite the clinical success of current Direct-Acting Antivirals (DAAs), the extreme genetic variability of HCV and the rapid emergence of drug-resistant mutations create a shifting therapeutic landscape. This constant viral evolution can compromise even highly effective regimens, necessitating continuous efforts to develop novel, resilient therapeutic strategies3.

Focusing on replicative enzymes, especially the NS5B polymerase, which is an RNA-dependent RNA polymerase crucial for the replication of the viral genome, is a well-established and highly promising strategy4. By inhibiting NS5B, the synthesis of viral RNA is blocked, which stops the creation of new virions and disrupts the replication cycle within hepato-cytes5. Unlike other targets, NS5B remains highly conserved across various HCV genotypes, making it an excellent target for broad-spectrum antiviral drugs6.The development of several key inhibitors has been driven by promising therapeutic applications. These inhibitors are categorized into nucleoside inhibitors (NIs), such as Sofosbuvir7, and non-nucleoside inhibitors (NNIs), including Dasabuvir and Becla-buvir8. Sofosbuvir acts as a chain terminator by being phosphorylated to its active triphosphate form, which subsequently competes with natural nucleotides for the NS5B active site9. The essential interactions involve hydrogen bonding with Asp225 and Ser282 residues, underscoring the importance of the phosphate group and modifications to the sugar moiety5,10. Conversely, NNIs bind to allosteric sites, such as the thumb or palm domains of the enzyme, inducing conformational changes that inhibit the initiation of replication11. While the active-site targeted NI Sofosbuvir is widely used and FDA-approved, NNIs offer a distinct mechanistic advantage by avoiding competition with high intracellular nucleotide concentrations. However, current NNIs face a critical limitation: their efficacy is highly dependent on the specific virus genotype, and they often exhibit a lower barrier to resistance compared to NIs12. Furthermore, many promising chemical scaffolds isolated in early-stage discovery fail in clinical trials due to poor bioavailability, high toxicity, or unfavorable pharmaco-kinetic behaviors13. Therefore, there is an urgent need to identify new chemical scaffolds that combine broad-spectrum pan-genotypic allosteric inhibition with optimized drug-like properties14.

To address these dual challenges of genotype-resistance and poor drug-likeness, this study focuses on the rational design of novel quinazolinone derivatives (QD) as potential NS5B inhibitors. Quinazolinone cores represent a privileged scaffold in medicinal chemistry, known for their structural versatility and diverse pharmacological profiles. Starting with the high resolution structural data of the NS5B-QD14 complex (PDB ID: 4JJU)15, we systematically explored in-situ modifications to optimize both binding affinity and safety profiles.

Traditional experimental screening of extensive chemical libraries is costly, time-consuming, and frequently overlooks pharmacokinetic flaws until late in development. To overcome this efficiency gap, our research utilized a rigorous, multi-step computational strategy designed to filter out weak candidates early. Initially, a quantitative structure-activity relationship (QSAR) model was constructed using a training set of QD with known experimental inhibitory activities15, linking chemical structure to thermodynamics via a Molecular Mechanics Poisson-Boltzmann (MM-PBSA) approach. Subsequently, a pharmacophore (PH4) model was derived from the bound conformations of the training set molecules to map the essential chemical features required for high-affinity binding. A virtual screening was then executed using this PH4 model to rapidly assess a virtual library of analogues, ensuring that newly proposed modifications retained or enhanced critical receptor interactions. Finally, the pharmacokinetic profiles of the selected analogues were rigorously evaluated to ensure favorable ADME properties, and their structural stability within the dynamic biological environment was verified through molecular dynamics simulations.

Ultimately, this integrated methodology directly addresses the limitations of current HCV therapies by streamlining the discovery of potent, structurally stable antiviral agents that possess both high pan-genotypic target affinity and optimized pharmacokinetic properties.

 

MATERIALS AND METHODS

 

Test and validation set of quinazolinone derivatives

The QSAR model was developed using 32 quinazolinone derivatives obtained from the literature15, with experimental inhibitory activities ranging from 60 nM to 64,000 nM. This substantial variation in activity enabled the construction of a robust QSAR model. A total of 24 quinazolinone derivatives were employed for the training and test set, while 8 were used for the validation set. Discovery Studio software16. was employed to partition the 32 quinazolinone derivatives into the test and validation sets. 

Model construction

Three-dimensional molecular models of the NS5B-QD1-24 complexes, the NS5B polymerase, and the QD1 to QD24 inhibitors were constructed from the NS 5B-QD14 complex (PDB ID: 4JJU, resolution: 1.91 Å)15 using Insight-II software17 via in situ modification of the co-crystallized ligand QD14 (Figure 1). 

Figure 2 illustrates the principal interactions of QD14 with the nearest active site residues. To ensure the reliability of the models, each structure underwent a systematic conformational analysis combined with a stepwise energy optimization protocol. The lowest-energy states were stabilized through a global minimi-zation of the system, following a standardized approach in the field of computer-aided drug design (CADD)18.

These computational protocols were implemented using the Discovery Studio16 and Insight-II17 software suites.

Molecular mechanics

We used a detailed atomic model and the charge parameters from the CFF force field to study the QD inhibitors, the NS5B protein, and their complexes. The molecular mechanics procedures implemented were executed according to the methodology described by Frecer and co-workers19,20.

Conformational search

The structures of the inhibitors in their unbound state were determined by relaxing their geometries from the enzyme-inhibitor complex (E: I) to the nearest local energy minimum. The conformational space was subsequently examined using the Monte Carlo method, permitting up to 50,000 iterations on all non-cyclic rotatable bonds, facilitated by the discover software17. For each inhibitor, 200 unique conformations were generated by applying random alterations to torsion angles by ±15° at a temperature of 5000 K, followed by energy minimization. During this process, a dielectric constant of (ε=80) was utilized to simulate the hydra-tion related screening effect. The conformer exhibiting the lowest total energy was ultimately selected and subjected to further minimization with a constant of (ε=4).  

Solvation free energy

The electrostatic component of the solvation Gibbs free energy (GFE), which includes the effects of ionic strength, was determined by solving the non-linear Poisson-Boltzmann equation21. This was accomplished using the Delphi module within Discovery Studio16. In this model, the solvent is represented as a continuous medium with high permittivity (ε=80), while the solute is considered a cavity with a low dielectric constant (ε=4). By applying finite difference methods on a cubic grid, along with physiological ionic strength and the parameters from the CFF force field, the GFE was derived as the reaction field energy21.

Binding Energy calculation and QSAR Model

The exhaustive methodology regarding the determination of binding affinity, quantified by the Gibbs free energy (∆Gcom) of complexation has been previously described in detail22. This calculation protocol, based on the equilibrium between solvated and bound states, relies on established theoretical frameworks for the evaluation of protein-ligand interactions23.   

Interaction energy

The quantification of interaction energies within the active site was performed using the CFF force field, in accordance with established computational proce-dures19. This calculation is based on the sum-ation of non-bonding contributions, including Van der Waals potentials and electrostatic interactions24.

Pharmacophore generation

The development of 3D-QSAR pharmacophore models was carried out using the bioactive conformations of the inhibitors extracted from the enzyme-ligand complexes. This modeling relied on the Catalyst HypoGen algorithm25, implemented within the Discovery Studio environment (Accelrys Inc., 2009). This methodological approach follows the validation protocols described by Bieri26.

ADME properties

To anticipate the pharmacokinetic profile of the new analogues, ADME descriptors were calculated using the QikProp program27 following the methodology of Duffy and Jorgensen28.

Virtual library

A virtual chemical library was constructed using the CombiLib protocol (Discovery Studio 2016) by systematically grafting R-groups onto the QD core, according to the methodology reported by Roméo29.

ADME-based library screening

The primary virtual chemical library was subjected to rigorous filtering based on drug-likeness descriptors to select only those QD analogues exhibiting an optimal pharmacokinetic profile30. This molecular curation process enabled the exclusion of compounds that did not comply with established criteria for oral bioavailability and membrane permeability31.

Pharmacophore-based library screening

The QD chemical library was refined through pharmacophore (PH4) screening within Discovery Studio. Only analogues displaying optimal alignment with the key features required for NS5B inhibition were retained32.

Prediction of inhibitory activity

The activity of the QD analogues selected after pharmacophore screening was then predicted using the QSAR scoring function, in accordance with established structure-activity relationship protocols33.

Molecular dynamics

The conformational stability of the QD22 ligand and its selected analogues was evaluated through 200 ns molecular dynamics simulations, as comprehensively detailed by Frecer and Miertus34. These simulations were conducted using the Desmond module from Schrödinger35, utilizing the OPLS 2005 force field.

MM-GBSA binding free energy calculation

Molecular dynamics simulations of the top analogues were employed to quantify variations in binding free energy using the MM-GBSA approach. For each inhibitor studied, the enzyme-inhibitor (E-I) complex with the lowest potential energy during the 200 nano-seconds was extracted from the trajectory and subse-quently subjected to energy minimization under the OPLS 2005 force field. The selection of this force field is justified by its established accuracy in evaluating solvation energies and structural parameters of heterocyclic ligands36. The variation in binding free energy is formalized by the following equation:

∆Gbind= ∆EMM + ∆Gsolv + ∆GSA  ….(Eq 1)

Where ∆EMM is the variation in molecular mechanics energy ∆Gsolv  is the solvation energy variation, and (∆GSA) is the surface area contribution. 

Statistical analysis

The QSAR model was constructed and rigorously validated using QSARINS software version 2.2.4-2019, which specifically implements the validation protocols required by the OECD (Organization for Economic Cooperation and Development)37-43. 

A univariate linear regression was established using the relative binding free energy ∆Gbind  as the descriptor to predict the experimental biological activity as the response. The dataset (n=32) was split into a training set (n=24) and a validation set (n=8).To ensure the model's robustness, internal consistency, and external predictivity, several statistical metrics were calculated across multiple dimensions of validation. Internal robustness was evaluated using the coefficient of determination, Leave-One-Out Cross-Validation (LOO CV), and Leave-Many-Out cross-validation (LMOCV), while potential over-fitting was monitored by comparing the Root Mean Square Error and Mean Absolute Error (MAE) against their cross-validated counterparts. External predictivity was simultaneously assessed using a validation set to determine the external squared correlation coefficient, the Root Mean Square Error of the validation set (RMSEext) and the Mean Absolute Error of the validation set (MSEext). Finally, to confirm that the observed correlation was not due to chance, a Y-randomization test was performed over 2000 iterations of Y-scrambling, utilizing the resulting random values and the  parameter which adjusts the original based on the randomization results to confirm the model's true statistical significance.

 

RESULTS AND DISCUSSION

 

Training and validation set

The training and validation set were constructed from thirty-two NS5B inhibitors (Table 1), designated as QD1–32, and sourced from benchmark studies15. The structural diversity and the heterogeneity of biological activities within this series ensure the robustness and predictive capacity of the developed QSAR model44. Using the Discovery Studio environment16, the 32 compounds were partitioned into a training set of 24 molecules and a validation set of 8 molecules. 

This balanced distribution guarantees an optimal coverage of the chemical space and activity ranges for both model calibration and external validation44.

QSAR Model

Single-descriptor QSAR

The relative Gibbs free energy was calculated for each NS5B-QDx complex based on the crystal structure of NS5B (PDB ID: 4JJU)45. Table 2 summarizes the calculated values of  as well as its various components for both the training and validation set.

A Hansch-type QSAR model was developed by creating a linear regression (Figure 3) that linked the experimental activities  with the  values. This model explains approximately 90% of the variance in experimental activities (Table 3). 

The statistical robustness of this correlation is confirmed by high R2 and R2XV  values, as well as a significant Fisher test (F=199.11, (Table 3). Furthermore, the gas-phase enthalpy was correlated with the pIC50exp values (Table 3). This correlation highlights the predominant contribution of interatomic interactions, which alone account for about 87% of the variation in biological activity.

Utilizing QSARINS software, we calculate additional key statistical parameters as detailed in Table 3, in compliance with OECD guidelines for regulatory QSAR models38,39,43. The QSAR model, can handle up to 30% of perturbation without affecting the LOOCV R2XV (0.88) and LMOCV Q2XV (0.88) values. Furthermore, the Y-scrambling method, which involves the random shuffling of responses, results in a low average R2Yscr value (0.04), markedly lower than the original model's value (0.90), indicating that the model is not subject to random correlations. Additionally, the model's predictive capability is assessed through low RMSE values for the training set (RMSE, 0.27), the cross-validation (RMSECV, 0.29), and the validation set (RMSEext, 0.46). The proximity of these RMSE values suggests enhanced generalizability of the model. 

Molecular dynamics

To evaluate the structural stability of the studied complexes, 200ns molecular dynamics simulations were performed for the most active ligand from the training set (QD22) and the top six analogues (1-2-5-45, 1-6-5-45, 4-1-3-45, 3-6-4-45, 5-1-6-45, and 5-3-2-45) 49. Examination of the root-mean-square deviations (RMSD) highlights good structural convergence (Figure 9, Figure 11). The reference complex (QD22) maintains an RMSD below 1.2 Å throughout the trajectory, while the analogues exhibit variations between 1.5 Å and 3 Å. These values  (1.5 ≤RMSD ≤3 Å) demonstrate that the introduced structural modifications do not affect the conformational equilibrium of the systems, as explained by Rana and co-workers50.

The time evolution of the radius of gyration (rGyr) served as an indicator of compactness. The (system/ complex) maintains an invariant value at 4.40 Å, reflecting high rigidity and dynamic stability. 

The analogues display rGyr values converging between 6.4 and 8.8 Å. These results show a displacement of the protein backbone (increasing the size of the pocket) according to Olivier Hucke and associates51, causing binding of the analogues with residues Pro 496, Leu 497, and Arg 498. This movement improves activity through hydrogen bonds and carbon-hydrogen bonds established with residues Arg 498, Leu 497, and Pro 496. 

Binding mode and interaction energy of new inhibitors

Analysis of the interaction diagram (Figures 10 and 12) illustrates the stable anchoring of the most active analogue within the allosteric pocket 2 of the NS5B polymerase. The oxoquinoline core forms key hydro-gen bonds with Tyr477 and Ser476, as described in the work of marina52, which are further reinforced by a π-sulfur interaction with Met423, thereby blocking the transition of the polymerase toward its active form53.

Linear regression analysis confirms the predictive accuracy of the selected model, displaying a slope close to unity (1.024) and a y-intercept close to zero (0.1419)42. The regression equation and its associated statistical parameters are as follows:

pIC50exp =1.0058 X pIC50pre + 0.0353; n=32, R2=0.85

R2XV =0.84, F-test=130.30; σ=0.34, α=95%

The correlation plot is illustrated in Figure 5E. Furthermore, the proper alignment of molecule QD22 with the pharmacophoric features of the PH4 model is shown in Figure 5D. Ultimately, this final PH4 model served as a predictive tool for pIC50 during the screening of a chemical library of QD analogues.

Virtual screening

The in silico screening of virtual combinatorial chemical libraries represents a choice strategy for identifying lead compounds (hits), as demonstrated by the work of Bléhoué et al.33.

Virtual library

A preliminary virtual library was constructed by systematically diversifying the quinazolinone (QD) scaffold at positions R1, R2, R3, and R4 (Table 5). The combinatorial introduction of 15 distinct substituents at each of the first three sites, and 50 substituents at the final site (R4), generated a chemical space of 168,750 structural analogues. All of these derivatives strictly maintain the substitution pattern of the most active reference inhibitor to preserve key interactions.

To optimize the drug-like profile of the molecules, a series of selection filters was applied. This protocol integrates Lipinski's “rule of five”, limiting molecular weight to 500 g/mol to ensure good oral bio-availability48. In the framework of this study, this filter was readjusted to also encompass the criteria required for potential intravenous administration. This rigorous filtration phase yielded a targeted, pharmacologically relevant chemical library optimized for subsequent virtual screening steps.

Virtual library screening

This library of 168,750 analogues was subjected to an in-silico evaluation to isolate molecular structures matching the geometric and electronic features of the 3D-PH4 Hypo1 pharmacophore model (NS5B inhibition). Following this filtering process, a total of 200 QD analogues satisfied the required pharma-cophoric characteristics. These preselected compounds (PH4 hits) were subsequently evaluated using the QSAR complexation model. The Gibbs free energy of NS5B-QD complex formation, its various energetic components, and the predicted values of the negative logarithm of the half-maximal inhibitory concentration (Table 6).

Substituent analysis of new QD analogues

In order to highlight the key substituents responsible for enhanced NS5B inhibition, an analysis was conducted on the 200 virtual hits aligned with the PH4 pharmacophore. Evaluation of the substituent distribution frequency at positions R₁, R₂ and R₃ (Figure 6) indicates that the introduction of small fragments at R₁, R₂ and R3 favors inhibitory activity, whereas at R4 activity is improved with the addition of bulky substituents. Notably, analogue 3-6-4-45 (Figure 7 and Figure 8) displays a predicted activity approximately 96-fold higher than that of the reference compound45.

In conclusion, this QSAR model is statistically rigorous and perfectly suited for predicting the affinity of new QD analogues sharing a comparable binding mode.

Molecular interaction analysis and binding mode

Examination of the NS5B–QD22 complex, modeled from the reference crystallographic structure (PDB ID: 4JJU)15, highlights a synergistic binding mode. The stability of the complex is primarily governed by a network of robust hydrogen bonds (His 475, Ser 476, Tyr 477) and π-π stacking interactions (Tyr 477, Trp 528), which ensure a rigid anchoring of the scaffold (Figure 4). Furthermore, the strong shape complemen-tarity observed with the residues (Leu 419, Leu 497, Leu 489, Ala 486, and Val 485) of the polymerase's hydrophobic pocket via π-alkyl interactions45  minimizes the Gibbs free energy, conferring a high predictive affinity to the QDx series. These interactions confirm that the QDx compounds optimally occupy the binding site, thereby effectively blocking the enzymatic activity of NS5B.

3D-QSAR pharmacophore model

A 3D-QSAR pharmacophore model for NS5B inhibition was designed using the bound conformations of all molecules from the training set and subsequently validated using the ligands from the validation set. This construction was performed as described in references26,46. To maximize the accuracy of the model and enhance its robustness, the biological activity uncertainty was reduced to 1.1. Among the ten best generated pharmacophores, the optimal model was selected (Table 4). It is characterized by a maximum cost difference (525.13), the highest correlation coefficient (0.86), the lowest RMSD value (3.17), and an acceptable configuration cost (9.62)16.

The CF3 group contributes through halogen bonds with Val485 and Ala486. Two additional hydrogen bonds stabilize the extremities of the molecule via Leu497 and Arg498, while alkyl/π-alkyl contacts (Leu497, Ile482, Leu419) and hydrophobic interactions of the chlorine atom with Arg501 and Trp528 complete the network.  Weak carbon-hydrogen bonds with Pro496 and His475 further refine the stabilization. This multivalent interaction network accounts for the high theoretical affinity of this analogue for NS5B.

To validate the obtained results, the relative MM-GBSA binding free energy calculated directly from the molecular dynamics simulation trajectories was compared to the relative Gibbs free energy of complexation54. The biological activities are within the same order of magnitude, as are each of the two thermodynamic quantities (Table 8). This convergence confirms the reliability of the computational approaches and allows us to propose these designed analogues as potential inhibitors.

Limitations of the study

Several limitations bound this work: the QSAR model's predictions are strictly restricted to the chemical space of its defined applicability domain; the validation metrics are inherently retrospective and may not perfectly mirror prospective experimental success; the statistical significance does not inherently guarantee a physical or biological causal mechanism; and the model's ultimate accuracy remains fundamentally capped by the baseline experimental noise and error present in the initial training data.

 

CONCLUSIONS

 

To characterize the inhibition of NS5B by quinazolinone derivatives (QDs), the X-ray crystal structure of the NS5B-QD14 complex was utilized. This approach demonstrated a strong correlation between the theoretical Gibbs free energies (GFE) and experimental inhibition data. These findings enabled the development of a 3D-QSAR pharmacophore (PH4) model, which was trained on 24 compounds and validated on 8 additional compounds. By examining the key interactions within the active site, a virtual chemical library of QD analogues was generated through targeted substitutions. This library was then filtered based on ADME criteria and mapped onto the PH4 model to ensure the bioavailability of the molecules. Finally, the properties of the top 15 virtual candidates were calculated using the QSAR model. These analogues represent strong candidates for the inhibition of NS5B polymerase. 

 

ACKNOWLEDGEMENTS

 

The authors gratefully acknowledge Prof. Paola Gramatica for granting a free license for the QSARINS software.

 

AUTHOR’S CONTRIBUTIONS

 

Yavo BU: conception, final drafting, writing original manuscript. Esmel AE: conception, investigations, technical, computational coaching, software, editing. Kouakou KJL: conception, investigations, computing, and editing. Kéita M: conception, intellectual input and final manuscript editing. Eugène M: intellectual input and final manuscript editing. Final manuscript was checked and approved by all authors. 

 

DATA AVAILABILITY

 

The associated author can provide the empirical data used to support the study's conclusions upon request.

 

CONFLICT OF INTEREST

 

None to declare. 

 

REFERENCES

 

  1. Spearman C. Wendy, Dusheiko GM, Hellard M, et al. Hepatitis C. Lancet 2019.
  2. Global hepatitis report 2024: action for access in low- and middle-income countries.
  3. Pawlotsky JM. Hepatitis C Virus Resistance to Direct-acting antiviral drugs in interferon-free regimens. Gastroenter 2016;151(1):70–86.https://doi.org/10.1053/j.gastro.2016.04.003
  4. Vaishali M. Patil Satya PG. Structural flexibility in HCV NS5B polymerase and molecular modelling of anti-HCV drugs . Curr C Bio 2018; 65–67.
  5. Moradpour D, Penin F, Rice CM. Replication of hepatitis C virus. Nat Rev Microbio 2007;5(6):453–463.https://doi.org/10.1038/nrmicro1645
  1. Powdrill MH, Bernatchez JA, Götte M, et al. Inhibitors of the hepatitis C virus RNS-dependent RNA polymerase NS5B. Viruses 2010; 2(10): 2169–2195.https://doi.org/10.3390/v2102169
  1. Gillian M. Keating. Sofosbuvir: A Review of its use in patients with chronic Hepatitis C. Drugs 2014;74:1127-1146.https://doi.org/10.1007/s40265-014-0247-z
  1. Sallamn M, Khalil R. Contemporary insights into hepatitis c virus: A comprehensive review. Microorganisms 2024;12:6. https://doi.org/10.3390/microorganisms12061035
  2. Brigham D, Narkewicz MR. Profile of Sofosbuvir and Velpatasvir Combination in the treatment of chronic Hepatitis C in children and adolescents: Current evidence. in therapeutics and clinical risk management 2024;20:1–7.https://doi.org/10.2147/TCRM.S326099
  1. Jácome R, Campillo-Balderas JA, Ponce de León S, et al. Sofosbuvir as a potential alternative to treat the SARS-CoV-2 epidemic. Sci Rep 2020;10(1):1–5.  https://doi.org/10.1038/s41598-020-66440-9
  1. Eltahla AA, Luciani F, White PA, et al. Inhibitors of the hepatitis C virus polymerase; mode of action and resistance. Viruses 2015;7(10):5206–5224. https://doi.org/10.3390/v7102868
  1. Di Marco L, Cannova S, Ferrigno E, et al. A comprehensive review of antiviral therapy for hepatitis C: The long journey from interferon to pan-genotypic direct-acting antivirals (DAAs). Viruses 2025;17:2.https://doi.org/10.3390/v17020163
  1. Sun D, Gao W, Hu H, Zhou S. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sin B. 2022;12(7):3049–62.  https://doi.org/10.1016/j.apsb.2022.02.002
  1. Yeung KS, Beno BR, Parcella K, et al. Discovery of a Hepatitis C Virus NS5B replicase palm site allosteric inhibitor (BMS-929075) advanced to Phase 1 clinical studies. J Med Chem 2017 May 25;60(10):4369–85.https://doi.org/10.1021/acs.jmedchem.7b00328
  1. Beaulieu PL, Coulombe R, Duan J, et al. Structure-based design of novel HCV NS5B thumb pocket 2 allosteric inhibitors with submicromolar gt1 replicon potency: Discovery of a quinazolinone chemotype. Bioorg and Med Chem Lette 2013;23(14):4132–4140. https://doi.org/10.1016/j.bmcl.2013.05.037
  1. Biovia DS. Discovery Studio Modeling Environment (2016). Dassault Systèmes.
  2. Accelrys Inc. (2005). Insight-II and Discover molecular modeling and simulation package (2005 ed.). Accelrys, Inc.
  3. Wu Z, Chen S, Wang Y, et al. Current perspectives and trend of computer-aided drug design: A review and bibliometric analysis. Int J Surg (London, England) 2024;110(6):3848–3878. https://doi.org/10.1097/JS9.0000000000001289
  4. Adama N, Alex AYJ, Bernard DAM, et al. Study by molecular docking of the interactions between Dihydroorotate Dehydrogenase and a series of inhibitors of pyrrole derivatives for the treatment of malaria. Asian J Chem Sci 2025;15(1):92–110. https://doi.org/10.9734/ajocs/2025/v15i1352
  1. Frecer V, Berti F, Benedetti F, et al. Design of peptidomimetic inhibitors of aspartic protease of HIV-1 containing -PheΨPro- core and displaying favourable ADME-related properties. J Mol Graph and Modelli 2008;27(3): 376–387. https://doi.org/10.1016/j.jmgm.2008.06.006
  1. Vergara PS, Marucho M. MPBEC, a Matlab program for biomolecular electrostatic calculations. Phy Commu 2016; 198, 179–194. https://doi.org/10.1016/j.cpc.2015.08.029
  2. Eugene M, Esmel A, Keita M, et al. Insight into binding mode of nitrile inhibitors of Plasmodium falciparum Falcipain-3, QSAR and Pharmacophore models, virtual design of new analogues with favorable pharmacokinetic profiles. SDRP J Comput Chem Mol Modelli 2017; 2(1). https://doi.org/10.25177/jcmp.2.1.5
  1. King E, Aitchison E, Li H, et al. Recent developments in free energy calculations for drug discovery. In Frontiers in Molecular Biosciences 2021;8. https://doi.org/10.3389/fmolb.2021.712085
  1. Wang B, Li L, Hurley TD, et al. Molecular recognition in a diverse set of protein-ligand interactions studied with molecular dynamics simulations and end-point free energy calculations. J Chem Inf and Model 2013;53(10):2659–2670.https://doi.org/10.1021/ci400312v
  1. Kumar H, Datusalia A K, Khatik GL, et al. Virtual screening of acetylcholinesterase inhibitors through pharmacophore-based 3D-QSAR modeling, ADMET, molecular docking, and MD simulation studies. Silico Pharmacol 2024;12(1).https://doi.org/10.1007/s40203-024-00189-1
  1. Bieri C, Esmel A, Keita M, et al. Structure-Based Design and pharmacophore-Based virtual screening of combinatorial library of Triclosan analogues active against Enoyl-Acyl carrier protein reductase of Plasmodium falciparum with favourable ADME profiles. Int J Molec Sci 2023;24(8). https://doi.org/10.3390/ijms24086916
  1. Schrödinger. (2019). No Title.
  2. Erin M. Duffy, William L. Jorgensen. Prediction of Properties from simulations: Free energies of Solvation in hexadecane, octanol, and water. J of the Ameri Chem Soc 2000;122(12):2878–2888.
  3. Roméo NG, Yvon BMB, Bernard DAM, et al. Molecular modeling and virtual screening: Application to computer-aided design of anticancer inhibitors of human dihydroorotate dehydrogenase. J Pharm 2025;37(8):52–80.https://doi.org/10.9734/jpri/2025/v37i87734
  1. Allangba KNPG, Keita M, Kre N’Guessan, et al. Virtual design of novel Plasmodium falciparum cysteine protease falcipain-2 hybrid lactone–chalcone and isatin–chalcone inhibitors probing the S2 active site pocket. J Enzy Inhibit and Med Chem 2019; 34(1): 547–561. https://doi.org/10.1080/14756366.2018.1564288
  1. Daniel F. Veber, Stephen R. Johnson, Hung-Yuan Cheng, et al. Molecular properties that influence the oral bioavailability of drug candidates. J Med Chem 2002; 45(12).
  2. Keïta M, Yaya Y, Bisseyou Yvon BM, et al. Molecular and thermodynamic modeling of the protein-ligand interaction. application to computer-assisted design of anti-competitive inhibitors of human histone deacetylase 2 (HDAC2). SDRP-J Comput Chem Mol Mod 2022; 605–630.https://doi.org/10.25177/jccmm.5.2.ra.10763
  1. Bléhoué IC, Koné M, Esmel EA, et al. Molecular modeling and virtual screening application to the computer-aided design of anticancer inhibitors with a favorable pharmacokinetic profile against e6 papillomavirus type 16. Universal J Pharm Res 2024;9(5).  https://doi.org/10.22270/ujpr.v9i5.1198
  1. Frecer V, Miertus S. Antiviral agents against COVID-19: structure-based design of specific peptidomimetic inhibitors of SARS-CoV-2 main protease. RSC Adv 2020;10(66):40244–63. https://doi.org/10.1039/D0RA08304F
  2. Schrödinger LLC, & D. E. Shaw Research. Desmond Molecular Dynamics System (3.6). Schrödinger, LLC 2018.
  3. Devleena SK, Joshua W, Yujie W, et al. Prediction of absolute solvation free energies using molecular dynamics free energy perturbation and the OPLS force field. J Chem Theo and Computat 2010; 6(5).
  4. Gramatica P, Cassani S, Roy PP, Kovarich S, Yap CW, Papa E. QSAR Modeling is not “Push a button and find a correlation”: A case study of toxicity of (Benzo-)triazoles on Algae. Mol Inform 2012 Dec 1;31(11–12):817–35.https://doi.org/10.1002/minf.201200075
  1. Gramatica P, Chirico N, Papa E, et al. QSARINS: A new software for the development, analysis, and validation of QSAR MLR models. J Comp Chem 2013; 34(24): 2121–2132. https://doi.org/https://doi.org/10.1002/jcc.23361
  1. Gramatica P, Cassani S, Chirico N, et al. QSARINS-chem: Insubria datasets and new QSAR/QSPR models for environmental pollutants in QSARINS. J Computat Chem 2014;35(13):1036–1044. https://doi.org/https://doi.org/10.1002/jcc.23576
  1. Gramatica P, Sangion A. A historical excursus on the statistical validation parameters for QSAR models: A clarification concerning metrics and terminology. J Chem Inf Model 2016 Jun 27;56(6):1127–31.https://doi.org/10.1021/acs.jcim.6b00088
  1. Gramatica P. Principles of QSAR Modeling: Comments and suggestions from personal experience. Int J Quant Structure-Prop Relation 2020;5(3):61–97.https://doi.org/10.4018/IJQSPR.20200701.oa1
  1. Gramatica P. Origin of the OECD Principles for QSAR validation and their role in changing the QSAR paradigm worldwide: An historical overview. J Chemom 2025 Mar 1;39(3):e70014. https://doi.org/10.1002/cem.70014
  2. Chirico N, Sangion A, Gramatica P, et al. QSARINS-Chem standalone version: A new platform-independent software to profile chemicals for physico-chemical properties, fate, and toxicity. In J Comput Chem 2021;42:20.https://doi.org/https://doi.org/10.1002/jcc.26551
  1. Muratov E N, Bajorath J, Sheridan R P, et al. QSAR without borders. Chem Soc Rev 2020; 49(11):3525–3564.https://doi.org/10.1039/d0cs00098a
  1. Kouman KC, Kouassi AF, Fagnidi YKH, et al. Virtual design of novel of orally bioavailable piperazine inhibitors of enoyl-acyl carrier protein reductase of Mycobacterium tuberculosis with favorable pharmacokinetic profiles. Universal J Pharm Res 2024; 9(5): 91-104. http://doi.org/10.22270/ujpr.v9i5.1216
  1. Kouassi AF, Kone M, Keita M, et al. Computer-aided design of orally bioavailable pyrrolidine carboxamide inhibitors of enoyl-acyl carrier protein reductase of Mycobacterium tuberculosis with favorable pharmacokinetic profiles. Int J Mol Sci 2015;16(12):29744–71.http://doi.org/10.3390/ijms161226196
  1. Accelrys Inc. Discovery Studio molecular modeling and simulation program. 2.5. Accelrys, Inc.; 2009.
  2. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 1997;23(1):3–25.https://doi.org/10.1016/S0169-409X(96)00423-1
  1. James C. Phillips, David J. Hardy, Julio D. C. Maia, et al. Scalable molecular dynamics on CPU and GPU architectures with NAMD. J Chem Phy 2020;153(4):044130.
  2. Rana N, Solanki P, Mehra R, Manhas A. Identification of natural compound inhibitors for substrate-binding site of MTHFD2 enzyme: Insights from structure-based drug design and biomolecular simulations. Chem Physics Impact 2025;10:100809. https://doi.org/10.1016/j.chphi.2024.100809
  3. Oliver Hucke, René Coulombe, Pierre Bonneau, et al. Molecular dynamics simulations and structure-based rational design lead to allosteric HCV NS5B polymerase thumb pocket 2 Inhibitor with picomolar cellular replicon potency. J Med Chem 2013;57(7). https://doi.org/10.1021/jm4004522
  4. Marina M. May, Dirk Brohm, Axel Harrenga, et al. Discovery of substituted N-phenylbenzenesulphonamides as a novel class of non-nucleoside hepatitis C virus polymerase inhibitors. Antivir Res 2012;95(2):182-191.https://doi.org/10.1016/j.antiviral.2012.04.008
  1. Biswal, B, K, Wang M, cherney MM, et al. Non-nucleoside Inhibitors binding to Hepatitis C Virus NS5B Polymerase reveal a novel mechanism of inhibitin. Journal of Molecular Biol 2006;361(1): 33-45.https://doi.org/10.1016/j.jmb.2006.05.074
  1. Gloua SM, Esmel AE, Zoro EGH, et al. Molecular modeling and virtual screening for computer-aided design of plant protoporphyrinogen IX oxidase (PPO) inhibitors. Sci Afr 2026; 31: e03208.https://doi.org/https://doi.org/10.1016/j.sciaf.2026.e03208