Introduction
Coronavirus disease (COVID-19) is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a virus identified in December 2019 in Wuhan, China. COVID-19 is currently recognized as a pandemic by the World Health Organization [1–3].
The first general genome of SARS-CoV-2 in Brazil (GenBank code: MT126808.1) was sequenced on March 1, 2020, and made publicly available at the National Center for Biotechnology Information (NCBI) on the following day [4]. The viral structure includes a spike glycoprotein (S protein), which is divided into S1 and S2 subunits. Together with ACE and TMPRSS2, S protein is important for SARS-CoV-2 entry into host cells [5]. The S1 region contains the receptor-binding domain (RBD), which interacts with angiotensin-converting enzyme 2 (ACE2) on the surfaces of target cells [6, 7]. This interaction contributes to the stabilization of the S2 domain, a key factor in enabling the fusion between viral and cellular membranes [5, 8]. Because S protein plays a key role in the pathogenesis of COVID-19, several studies have examined the use of monoclonal antibodies targeting this protein; these antibodies’ promising neutralizing effects on viral entry into cells in vitro support their application potential in COVID-19 prevention [9, 10].
No specific treatments are available for COVID-19. However, since the beginning of the pandemic, scientists have sought effective vaccines or treatments with new compounds or repositioned FDA approved drugs [11]. Given the structural-functional relationships between the S protein of SARS-CoV-2 and ACE2—and the importance of the S protein in attachment and entry, resulting in the replication cycle—the S domain might be a promising antiviral drug target. In the present study, we used docking models and conducted an in silico analysis of the molecular interactions between FDA approved drugs and the RBD encoded by the Brazilian S Omicron genome sequence, to develop a strategy for functional evaluating drug repositioning.
Methods
Sequence alignment and modeling
Homology modeling is widely applied to create a reliable protein structure by using a protein’s own amino acid sequence [12]. The complete genome sequence of the Omicron variant of SARS-CoV-2 in Brazil (ON241705.1) was downloaded from the NCBI nucleotide bank (https://www.ncbi.nlm.nih.gov/). The SWISS-MODEL web server (https://swissmodel.expasy.org) was used to build the SARS-CoV-2 S protein RBD model by using “Alignment Mode” [13].
The structure obtained (PDB ID: 6W41) was used as a model, showing 99.55% similarity to the RBD homolog sequence. After generating the RBD model using SWISS-MODEL, we evaluated its three-dimensional quality and structural integrity using the Structure Analysis and Verification Server (SAVES v6.0) from the University of California at Los Angeles [14, 15]. To verify the model’s validity, we used PROCHECK [16], Verify 3D [17], PROVE [18], and ERRAT [19] software. After model validation, we performed structural minimization in UCSF CHIMERA v1.14 software by using the AMBER ff14SB force field [20] and PROSA [21].
Preparation of ligands
ChemSpider (http://www.chemspider.com/) is a chemical database containing more than 67 million structure-based chemistry records [22]. The ligands umifenovir (CSID: 116151), darunavir (CSID: 184733), lopinavir (CSID: 83706), ritonavir (CSID: 347980), remdesivir (CSID: 58827832), pirfenidone (CSID: 37115), and oseltamivir (CSID: 58540) were obtained from this virtual repository in mol format. All structures were calculated and geometrically optimized with the MMFF94s force field in the GAMESS package [22, 23].
Molecular docking
Before docking, the receptor protein was carefully prepared to ensure the accuracy of the experiments. Active pockets and cavities within the protein structure were identified with the CASTp web server (http://sts.bioe.uic.edu/castp/index.html) with a probe radius of 1.4 Å, the standard value for detecting potential active sites [24].
Docking experiments were performed with AutoDock Vina and AutoDock Tools [25]. During preprocessing, all water molecules and non-essential residues in the RBD of the SARS-CoV-2 Omicron variant were removed to eliminate any interference with the docking process. Polar hydrogen atoms were added to the protein structure to facilitate the formation of hydrogen bonds between the receptor and the ligands. Charges were assigned with the Gasteiger charge model to ensure accurate electrostatic calculations.
The grid box for the RBD was configured to fully encompass the binding site, with the following coordinates: X = −40.0, Y = −39.978, and Z = −6.261. The grid box dimensions were set to X = 80, Y = 80, and Z = 94, thus providing sufficient space for ligands to enable effective exploration of various binding conformations.
The binding affinity of the ligands was assessed according to their Gibbs free energy (ΔG). A threshold of ΔG ≤ −8.0 kcal/mol was used to identify candidates with strong binding potential for SARS-CoV-2 treatment. Binding interactions were visualized and analyzed in both 3D and 2D formats with PyMOL v2.0 [26] and LigPlot+ v2.2 [27], respectively. These visualizations highlighted key contacts, including hydrogen bonds and hydrophobic interactions.
Molecular dynamics
All molecular dynamics (MD) simulations were conducted with GROningen MAChine for Chemical Simulations (GROMACS) version 5.1.2 [28]. The OPLS-AA/L force field and the TIP3P water model integrated into GROMACS were used for the MD simulations [29]. The peptide topology file was generated with GROMACS, and the ligand topology was obtained with LigParGen. To satisfy minimum image conventions, the system was initially placed in a cubic simulation box with a 2 nm distance between the protein complex and the box boundaries. All bond lengths in proteins and ligands were constrained with the LINCS algorithm, whereas water molecules were constrained with the SETTLE algorithm [30, 31]. A total of 30,000 water molecules were added to a cubic simulation box containing the unbound RBD-Omicron structure. Each system was energy-minimized with the steepest descent algorithm to ensure optimal stability. Short-range and long-range non-bonded interactions were calculated with dual cutoff values of 0.9 nm and 1.4 nm, respectively. The leapfrog algorithm was used to integrate the equations of motion with a time step of 2 fs, and the neighbor list was updated every five steps. Long-range electrostatics were computed with the particle mesh Ewald method with a Fourier grid spacing of 0.15 nm [32].
Periodic boundary conditions were applied in all three spatial dimensions. System equilibration was performed in two main stages: the system was gradually heated to 300 K under constant volume (NVT ensemble) for 10 ns with the v-rescale thermostat, followed by a constant pressure (NPT) ensemble for 5 ns, during which the complexes (RBD-Omicron/ritonavir and RBD-Omicron/lopinavir) were constrained, thus allowing the solvent molecules to relax around them. An additional 10 ns NPT equilibration was performed, during which the constraints on the complexes were removed. Throughout the equilibration phases, the temperature and pressure remained close to target values. The equilibrated systems were then subjected to production MD simulations for 100 ns, while constant temperature (300 K) and pressure (1 bar) were maintained. We ensured reproducibility by conducting two independent simulations with different initial velocities and equilibration times. Root-mean-square deviation (RMSD) and root-mean-square fluctuation (RMSF) values were calculated from the MD simulation trajectories to assess system stability and residue flexibility [33].
Results and discussion
On the basis of the sequence identity between SARS-CoV-2 and the RBD, we used SWISS-MODEL to build a valid and high-quality model with 93.85% similarity to the structure of 6W41. Figure 1A shows the superposition results. The corresponding regions in the SARS-CoV-2 genome were identified according to the sequence identity and similarity score. Figure 1B shows the amino acid sequence between the constructed structure and the model example structure (223 amino acids). The 14 mutations (yellow and green) can be seen in the sequence of the RBD of the Brazil-2022 Omicron strain compared with the original strain, Wuhan-2019. As a result of the high sequence similarity, the homology models revealed an unexpectedly conserved overall architecture. The overlap between RBD-BR and the RBD (6W41) of SARS-CoV-2 indicated high structural similarity, with RMSD values of 0.35Å. In addition, the average value of the classifier QMEANDisCo [34] Global was 0.83 ± 0.05 ( Figure 1C ). The results confirmed the high quality of the model, for which the default score ranged from 0 to 1 [35].

Results of structural modeling according to Omicron RBD structural homology. (A) The Omicron RBD model, built with Swiss-Model is in purple, and the superimposed structure (6W41.1.C) is in yellow. (B) Comparison of sequences of the Brazilian built RBD model (Omicron-RBD) and the structure used in the homologous construction (6W41.1.C). (C) Structural validation chart of the classifier QMEANDisCo, referring to RBD-Omicron homology. (D) Ramachandran plot of the RBD-Omicron structure modeled in 3D. (E) Energy prediction results, as compared with the PROSA database structure. (F) Z-score of the RBD-Omicron structure.
Subsequently, the modeled structure was validated through PROCHECK, Verify 3D, and PROSA II analyses. PROCHECK analysis predicted that the hypothetical model had 89.6% of residues in the favored region, 9.1% of residues in the additionally allowed region, 1.2% of residues in the generously allowed regions, and 0% of residues in the disallowed regions. The quality factor of the residues from the Omicron RBD model, evaluated with Verify3D, indicated that 98.26% of the residues had a favorable average 3D-1D score ≥ 0.2, suggesting high compatibility of the atomic model (3D) with the amino acid sequence (1D). The ERRAT Ramachandran plot predicted that the final model had 97.8% of residues in favored regions, 2.2% of residues in allowed regions, and 0.0% outlier residues ( Figure 1D ). A comparative analysis was performed with the ProSA II web algorithm, which predicted that the final structure had a Z-score of −5.91 and therefore was also within the range of scores established for proteins of similar size, with NMR quality ( Figure 1E and F ).
The viral entry pathway in host cells might be an effective treatment target. In Figure 2 , we predicted the conformations of seven therapeutic drugs in complex with the Omicron RBD model, which blocked binding to the putative receptor, human ACE2. The tested drugs in this study were umifenovir, a broad-spectrum antiviral agent used in influenza prophylaxis and treatment, which also has in vitro antiviral effects against SARS-CoV [36, 37]; oseltamivir, a neuraminidase inhibitor used to treat influenza [38]; remdesivir, a nucleotide analog prodrug used to treat Ebola virus [39, 40]; the HIV protease inhibitors darunavir, lopinavir, and ritonavir [41]; and the antifibrotic agent pirfenidone, which is effective against idiopathic pulmonary fibrosis [42, 43].

Histogram of binding affinity results (ΔG) between drugs and the modeled RBD. The dashed line shows the limit value of −8 kcal/mol.
Umifenovir has been reported to have antiviral activity in vitro [35]. A pilot clinical trial conducted in 36 patients in Wuhan city, using reverse transcriptase-polymerase chain reaction, suggested a potential reduction in both viral load and mortality rates [44]. In addition, a clinical trial using a combination of lopinavir and ritonavir has demonstrated promising results [45]. A phase 4 study is ongoing [46]. However, our molecular docking analysis identified a binding score of −5.7 kcal/mol, indicating moderate affinity for the RBD. This suggests that the positive results observed in initial clinical trials may not be directly related to interactions with this domain. Instead, they could potentially involve immunological mechanisms, such as modulation of the immune response, which require further investigation to be fully understood.
Oseltamivir, a neuraminidase inhibitor commonly used as an anti-influenza drug, has recently been suggested as potentially effective against SARS-CoV-2. This drug is currently being evaluated in a phase 3 randomized controlled trial; however, limited evidence is available regarding its mechanism of action [47]. In our molecular docking analysis, oseltamivir exhibited a binding score of −4.9 kcal/mol with the RBD, indicating a weaker binding affinity compared to other compounds. A previous docking study demonstrated its interaction with both papain-like protease (PLpro) and main protease (Mpro), highlighting these as possible targets. Notably, oseltamivir alone showed a lower binding score (−4.65 kcal/mol) compared to the lopinavir/ritonavir combination, which exhibited a stronger interaction with a score of −8.3 kcal/mol [48]. However, in a separate in silico study assessing various SARS-CoV-2-targeting drugs, oseltamivir did not demonstrate significant anti-SARS-CoV-2 potential [49].
We also evaluated remdesivir, a nucleotide analog that inhibits RNA polymerase, which demonstrated a binding affinity of −6.0 kcal/mol. When tested against MERS-CoV, remdesivir was found to interact with the non-structural protein 8 (Nsp8) and Nsp12 domains in vitro [50]. Additionally, it has been confirmed that remdesivir associates with both exonuclease (ExoN) and RNA-dependent RNA polymerase (RdRp or Nsp12) [51]. RdRp is a viral enzyme essential for the replication of RNA viruses, including SARS-CoV-2. Elfiky [52] reported that remdesivir binds tightly to RdRp with a binding energy of −7.6 kcal/mol. At the ACE2 binding site, remdesivir exhibited a lower binding energy score of −5.62 kcal/mol [53]. Furthermore, molecular docking studies have identified efficient binding energy of remdesivir toward other targets, including Nsp10, Nsp16, Mpro, the S protein, and other components in their unbound forms [49].
In an animal model of infection, remdesivir has also shown anti-SARS-CoV-1 activity mediated by the viral polymerase and the proofreading exoribonuclease [54]. Additionally, remdesivir has been found to efficiently inhibit SARS-CoV-2 in rhesus monkey cell culture [44, 55]. Ongoing clinical trials have demonstrated remdesivir’s effectiveness against COVID-19 [55] (phase 3 NCT04292730, March 3, 2020; NCT04292899, March 3, 2020; and NCT04401579, March 3, 2020). On the basis of these findings, the interaction between remdesivir and RBD does not appear to inhibit the receptor-binding domains of SARS-CoV-2. However, its crucial role in RNA synthesis, combined with its ability to interact with ExoN (an enzyme located in the N-terminal domain of Nsp14 that reduces the rate of RNA excision), might help explain its observed efficacy in humans.
The interactions with the best binding affinity (the most negative binding energy values) were observed to fall below the threshold of the gray score limitation line (−8.0 kcal/mol). Drugs with such scores demonstrated strong binding to the RBD of the S protein, which plays a key role in the molecular replication of SARS-CoV-2. Consequently, these drugs might serve as potential candidates for inhibiting viral replication processes and complementing the ongoing use of other drugs currently in clinical trials that have shown promising results [56]. Among the tested drugs, lopinavir and ritonavir showed the best results in RBD viral protein docking (−9.8 kcal/mol and −8.9 kcal/mol, respectively). Darunavir, another HIV protease inhibitor, was not among the best docked drugs (−5.9 kcal/mol).
Figure 3 shows the best results for molecular docking simulation, residue interaction analysis, and MD simulation. The residues involved in hydrophobic interactions with the RBD structure are represented by dashed semicircles around the structure. The H bonds are shown as green dashed lines. Lopinavir ( Figure 3C ) showed 11 interactions: hydrophobic interactions involving ILE8, VAL9, ARG10, PHE11, LEU72, LEU199, LEU200, ALA202, PRO203, ALA204, and PHE 223, and one conventional hydrogen bond to CYS73. Ritonavir ( Figure 3D ), showed ten interactions: hydrophobic interactions involving VAL9, PHE11, ASP71, LEU72, CYS73, HIS201, PRO203, ALA204, LYS210, and PHE 223, and three conventional hydrogen bonds to residue ARG10.

Results of docking and molecular dynamics evaluations. (A) Interaction with lopinavir in the 3D model. (B) Interaction with ritonavir in the 3D model. (C) Lopinavir in the 2D model. (D) Ritonavir in the 2D model. (E) RMSD plot of the complexes and isolated RBD. (F) RMSF of the RBD-lopinavir complex. (G) RMSF of the RBD-ritonavir complex.
The RMSD quantifies the average positional deviation of Cα atoms relative to the initial structure over the course of the simulation, thereby providing insights into the system’s structural stability. RMSD values ≤ 2.00 Å are generally considered to indicate a stable protein structure during MD simulations [57]. As shown in Figure 3E , the RMSD profiles of the RBD, RBD-lopinavir complex, and RBD-ritonavir complex demonstrated stabilization within the 100 ns simulation timeframe.
Although the RMSD values for the complexes were slightly higher than those for the isolated structure, this difference does not indicate substantial instability, but instead reflects the conformational adjustments and flexibility induced by ligand binding, which are typical and expected phenomena in protein-ligand interactions. This flexibility might enhance the protein’s ability to accommodate and stabilize ligands within the binding site.
RMSF analysis indicates residue fluctuations and flexibility during simulation, wherein greater fluctuations are associated with lower stability [58]. This parameter is essential for exploring the roles of individual amino acids in the stability of any binding protein complex. The RMSF of each amino acid in the RBD-lopinavir ( Figure 3F ) and RBD-ritonavir ( Figure 3G ) complexes showed similar fluctuations and only minor variations. Residue CYS073 of the RBD-lopinavir complex was identified in the graph for its fluctuation of 7.23 Å, because of the hydrogen interaction shown in Figure 3C , thus confirming the findings from the residue interaction analysis. Figure 3G shows the highest fluctuation peak for an ARG010 oscillation, thereby also confirming the hydrogen bond observed in Figure 3D .
Because lopinavir and ritonavir are protease inhibitors, several drug repurposing studies have investigated lopinavir and ritonavir alone or in combination. In an in vitro evaluation, lopinavir has demonstrated promising results [59], and in combination with ritonavir, it has shown inhibitory effects toward SARS-CoV-2 [60]. Computational studies have focused on evaluating these drugs against proteases present in SARS-CoV-2. Studies using docking and MD have indicated strong affinity of these drugs toward SARS-CoV-2 Mpro, with a free energy of binding of −10.89 kcal/mol (lopinavir) and −14.93 kcal/mol (ritonavir) [61]. In addition, the combination of lopinavir, ritonavir, and oseltamivir improved the free energy values of binding toward Mpro, thus making the system more stable [62]. Efficient binding energy has been observed between ritonavir and Nsp10, Nsp16, and S protein, and between lopinavir and Mpro/3CLpro, the ADP ribose phosphatase Nsp3, Nsp9, and the ACE2 receptor protein [49].
For clinical use, the lopinavir/ritonavir combination is particularly effective because ritonavir, a strong reversible inhibitor of cytochrome P450 3A4, reduces lopinavir’s metabolism, ensuring higher and more sustained plasma levels [63, 64]. Currently, lopinavir/ritonavir are under phase 3 clinical trials (NCT04303299). However, clinical trials with protease inhibitors have not demonstrated sufficient positive interactions for activity against infection. Darunavir and lopinavir/ritonavir might not improve clinical outcomes in the treatment of mild/moderate COVID-19 [64], and darunavir in combination with cobicistat has not achieved clinical improvement over the control [65].
Conclusion
The present study demonstrated that the protease inhibitors lopinavir and ritonavir act by binding the RBD of S protein, an extremely important region for viral entry into cells, according to modeling based on the SARS-CoV-2 genome found in Brazil. This work may support future studies evaluating the mechanisms of action of these drugs against SARS-CoV-2. In addition, it is essential to highlight the significance of research focused on the rational design of drugs, as these molecules can serve as foundational models for further investigations. Understanding potential mechanisms of action at the molecular level is crucial for advancing therapeutic strategies.