The vital target (essential for host–pathogen interaction, replication, and pathogenesis) sequences within the HIV mechanism were retrieved from UniProt (https://www.uniprot.org/). The vaccine protein must have strong immunological properties and be non-allergenic to confirm a potent immune response (29). These retrieved sequences were further subjected to the antigen and allergen assessment via VaxiJen v2.0 (https://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html) (30) considering virus as a target and a threshold value of 0.4 and the AllerTOP v.2.0 (https://www.ddg-pharmfac.net/allertop_test/) (31) server. The VaxiJen server is mainly based on alignment-based prediction methods, while the AllerTOP server is alignment-free and grounded on the target’s physicochemical properties.

Two subsequent servers—ABCpred (http://crdd.osdd.net/raghava/abcpred/) (32), which utilized the artificial neural network, and BepiPred 2.0 (http://tools.iedb.org/bcell/) grounded on the sequence features of the antigen (33), available at IEDB—have different algorithms to detect more potential linear B-lymphocyte (LBL) epitopes considering collected sequences as input with default parameters. For peptide vaccines, recognizing B cells is crucial, as their receptors recognize peptides to trigger an effective immune response (29). However, the epitopes were further considered based on their presence in both servers and examined via VaxiJen v2.0 (30), AllerTOP v.2.0 (31), and ToxinPred (https://webs.iiitd.edu.in/raghava/toxinpred/) along with default parameter (34) servers.

CD8⁺ T lymphocytes recognize MHC I epitopes. When a cell is infected or has aberrant proteins (such as in viral infections or cancer), MHC I molecules present these peptides on the cell surface, prompting CD8⁺ T cells to kill the infected or abnormal cells (35–37). On the other hand, CD4⁺ helper T cells recognize MHC II epitopes. Antigen-presenting cells (APCs) internalize and process foreign antigens, presenting peptides on MHC II molecules to activate CD4⁺ T cells, which then help coordinate the broader immune response (38, 39). The MHC I and MHC II within the targets were identified using Tepitool (http://tools.iedb.org/tepitool/), which computes the epitopes based on seven prediction methods (IEDB recommended, consensus, NetMHCIIpan, NN-align, SMM-align, Sturniolo, and the combinatorial library method) (40). For MHC I, 27 and MHC II, 7, the most frequent alleles with the restricted 9- and 15-mer length were selected, and all other IEDB-recommended parameters were selected (24, 29, 40). Furthermore, the immune assessment was done similarly to the abovementioned one to screen out the potential epitopes.

To formulate a mutation-proof vaccine, the designed vaccine should be highly effective in both mutated and non-mutated forms (29, 41). The available sequence concerning each target was collected from UniProt (https://www.uniprot.org/). These sequences were subjected for multiple sequence alignments via Clustal Omega (https://www.ebi.ac.uk/jdispatcher/msa/clustalo), which is based on the seeded guide trees and the HMM technique (42), and the variable amino acid was visualized and collected using the JalView (43) software. These variable amino acids were further mapped with the final selected B- and T-cell epitopes to incorporate mutation.

The highly antigenic score followed by non-allergenic and non-toxic-based LBL, MHC I, and MHC II epitopes were selected from each target for the vaccine formulation, leading to a robust immune response against the infection. These epitopes were joined via different subsequent linkers (EAAAK, AAY, KK, and GPGPG) (21, 41). Furthermore, to enhance, activate, and purify, the adjuvant, PADRE, and His-tag were also attached at the N and C terminals of the vaccine construct. In contrast, His-tag was attached using the RVRR linkers (5, 7, 11, 21). Moreover, considering combination, six different vaccines were constructed to identify additional potential combinations with high antigenic properties (score). However, the adjuvant (beta-defensin), PADRE at the N, and His-tag at the C terminal were kept in different distinct vaccine constructs (44).

Moreover, the EAAAK offers an extended, uncharged spacer that can reduce steric hindrance in the region, AAY enhances the immunogenicity and improves pathogen-specific immunity while reducing junctional immunogenicity, KK linkers enhance solubility and are crucial proteases required for antigen processing, GPGPG linkers will aid to avoid aggregation and sustain flexibility, and His-tag is vital for the recognition and separation and facilitates efficient purification (7, 29, 44–46). The antigen and allergen predictions were used similarly to those mentioned above to identify vaccine candidate combinations with optimal immunological and antigenic properties. The combination with the highest antigen score was analyzed for its physicochemical activity via the ProtParam server (https://web.expasy.org/protparam/) (47), considering default parameters. The selected vaccine combination also underwent solubility analysis via Protein-sol (https://protein-sol.manchester.ac.uk/) (48), which is based on weighted scores considering default parameters.

The selection of potential must be validated based on its population coverage, which can be crucial for vaccine development and helpful for most of the world’s population (29). The final MHC I and MHC II epitopes with their restricted alleles were utilized for the analysis via population coverage (http://tools.iedb.org/population/) (49), which estimates the fraction of responders to epitopes with known MHC restrictions.

The variable positions identified through multiple sequence alignment were mapped onto the selected epitope to introduce variability and design a mutated epitope to formulate a mutated vaccine that can be helpful in combating multi-strain. The mutated vaccine was constructed, and its immune and physiological assessments were performed similarly to those of the non-mutated vaccine.

The SOPMA (based on the homology modeling) (https://npsa-prabi.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma.html) (50) and PSIPRED [based on machine learning (ML)] (http://bioinf.cs.ucl.ac.uk/psipred/) (51) were employed to examine the secondary structure of non-mutated and mutated vaccine construct following the default parameters. However, structure was modeled via the Robetta (https://robetta.bakerlab.org/) (52) server based on deep learning methods using RoseTTAFold. These models were enhanced via the GalaxyRefine (https://galaxy.seoklab.org/cgi-bin/submit.cgi?type=REFINE) (53) server, and the most promising enhanced models were further examined for their structure quality validation via PROCHCK (https://saves.mbi.ucla.edu/) (54) and ProSA-web (https://prosa.services.came.sbg.ac.at/prosa.php), which is grounded on the statistical analysis following the available structure (55).

Discontinuous epitopes are crucial for encoding the immune system’s specificity and complexity in responding to infectious agents, leading to more robust and protective immune responses (36). Therefore, the presence of these epitopes within the non-mutated and mutated vaccine was examined via Ellipro (http://tools.iedb.org/ellipro/) (56), which is grounded on geometrical properties of structure, considering the vaccine model structure.

Potent vaccines must be able to bind with the receptor to activate an immune activity. Therefore, the formulated vaccine (non-mutated and mutated) was docked with the TLR via the ClusPro (https://cluspro.org/login.php) (57) webserver, which utilized the PIPER docking algorithm following the default parameters, whereas the TLR3 structure was collected via the Protein Data Bank (PDB) (ID: 1ZIW) (https://www.rcsb.org/) database. The obtained docked complexes were examined, and the most potent complexes were selected based on their lowest negative energy, demonstrating strong binding. The binding affinities of complex chosen were computed via the PRODIGY (https://rascar.science.uu.nl/prodigy/) (58) sever, and their interaction was visualized through the PDBsum (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/) (59) and PyMOL.

To examine the docked complex’s stability (vaccine with TLR), the Desmond software on an Acer workstation with Ubuntu 20.04 was used (60). The OPLS-2005 Force field was employed to generate the coordinates and topology file of the vaccine and TLR complex to define bonded and non-bonded interactions. The system was prepared, solvated (in the TIP3P model), and further neutralized to mimic the physiological condition via Na+ and Cl− counter ions with 0.15 M salt concentration. Furthermore, the simulations were carried out at 300 K temperature and 1.0325 bar pressure for 100 ns, and the system was minimized and relaxed using the default protocol considering all other criteria that were earlier described (23, 60–63). Furthermore, the trajectory file was examined by root mean square deviation (RMSD) and root mean square fluctuation (RMSF) to evaluate the system’s stability.

The immune activity produced via vaccine (non-mutated and mutated) was analyzed via C-ImmSim (https://kraken.iac.rm.cnr.it/C-IMMSIM/index.php) (64), which employs an ML algorithm. This server assesses the host’s immune activity and the ensuing vaccine administration. Default parameters were used following the adjustment based on previously reported data corresponding to the vaccine construct sequence. Additionally, time steps were modified to reflect the administration of three doses at 1, 84, and 168, with 1,050 set as the simulation step, while all other parameters remained the same (5, 21, 65).

The formulated vaccine (non-mutated and mutated) must have a high expression level for a robust response. Therefore, the constructed sequence was optimized via the VectorBuilder (https://en.vectorbuilder.com/tool/codon-optimization.html) server, considering E. coli K12 with default parameters. The Codon Adaptation Index (CAI) and GC% should be 0.8–1.0 and 30%–70% for the maximum expression, respectively (25, 66). Furthermore, the optimized sequence was incorporated and cloned in pET-28a (+) via SnapGene (https://www.snapgene.com/) software, considering a specific restriction site as previously reported (5, 7, 11).

The selected proteins, envelope glycoprotein (Q75008), protease (Q75002), reverse transcriptase (Q75002), and integrase (Q75002), were retrieved from the UniProt database, which is a part of the human immunodeficiency virus type 1 group M subtype C (isolate ETH2220), and are crucial in the infection mechanism (11, 67). The immune assessment of the target sequence demonstrated ( Table 1 ) that the required properties can be utilized for vaccine formulation.

The crucial B-cell epitope within targets was identified via ABCpred (32) and BepiPred 2.0 (33). Via the ABCpred server, 87 envelope glycoprotein ( Supplementary Table 1 ), 9 protease ( Supplementary Table 2 ), 56 reverse transcriptase ( Supplementary Table 3 ), and 28 integrase ( Supplementary Table 4 ) epitopes, and simultaneously via BepiPred, 28 envelope glycoprotein ( Supplementary Table 5 ), 4 protease ( Supplementary Table 6 ), 20 reverse transcriptase ( Supplementary Table 7 ), and 9 integrase ( Supplementary Table 8 ) epitopes were predicted. Moreover, 25 envelope glycoprotein, 4 protease, 24 reverse transcriptase, and 13 integrase epitopes were selected to screen out the more precise assessments, which overlapped in both ( Supplementary Table 9 ). The immune evaluation of these final epitopes revealed that several epitopes have potential, having antigen, non-allergen, and non-toxic features, and the epitopes with high antigen scores from each target ( Supplementary Table 9 , highlighted in blue) were selected for vaccine formulation as in Table 2 .

The MHC I and MHC II epitopes were identified within the targets via Tepitool (40), considering the most frequent alleles (29). The MHC I assessment revealed 238 envelope glycoprotein ( Supplementary Table 10 ), 24 protease ( Supplementary Table 11 ), 170 reverse transcriptase ( Supplementary Table 12 ), and 76 integrase ( Supplementary Table 13 ) epitopes. Simultaneously, the MHC II assessment revealed 80 envelope glycoprotein ( Supplementary Table 14 ), 12 protease ( Supplementary Table 15 ), 61 reverse transcriptase ( Supplementary Table 16 ), and 32 integrase ( Supplementary Table 17 ) epitopes. The immune assessments of the epitope in MHC I and MHC II revealed several leading immunodominant properties, as shown in Table 3 . Furthermore, one epitope with many covering alleles and a high antigenic score ( Table 3 ) from each respective target was selected for vaccine formulation, as in Table 4 .

To compute the variability of amino acids across different variants, the total reviewed sequences concerning each target were retrieved from UniProt, and their MSA was accomplished via Clustal Omega (42). The MSA was visualized via the JalView (43) software, which revealed several variable positions across the variant ( Supplementary Figures 1 – 4 ). In the case of the B-cell epitope, a total of 38 amino acids from envelope glycoprotein, 29 from protease, 16 from reverse transcriptase, and 9 from integrase were found and mapped ( Supplementary Table 18 ) with the selected final epitope ( Table 2 ), whereas 73 amino acids from envelope glycoprotein, 8 from protease, 21 from reverse transcriptase, and 12 from integrase for the combined MHC I and II were found and successfully mapped ( Supplementary Tables 19 – 22 ) with the selected epitope ( Table 4 , non-mutated vaccine formulation). These mapped amino acids were further incorporated (highlighted in red), and the variability was introduced in the selected non-mutated B- and T-cell epitope ( Tables 2 , 4 ). Furthermore, the mutated epitope ( Supplementary Tables 19 – 22 ) concerning to non-mutated epitopes were examined for antigen, allergen, and toxicity assessment, similar to those mentioned for non-mutated epitopes, and several potential epitopes were found to have antigenic, non-allergenic, and non-toxic properties ( Supplementary Tables 19 – 22 ). Among the potential epitopes, the epitopes with high antigenic scores ( Supplementary Tables 19 – 22 , highlighted in blue) were further selected for mutated vaccine formulation.

Among the predicted epitopes, four LBL ( Table 3 ), four MHC I, and four MHC II ( Table 4 ) were selected based on their high immunodominant activity for the non-mutated vaccine formulation. In contrast, four LBL, four MHC I, and four MHC II mutated epitopes concerning the non-mutated vaccine, based on the introduced variability having high antigenic scores, were used for mutated vaccine formulation, as in Table 5 . These selected epitopes were joined via EAAAK, AAY, KK, and GPGPG linkers to attain the most immunodominant combination; six distinct non-mutated vaccines were constructed considering the selected epitope and different linkers, adjuvants, and other essential attributes. Moreover, the adjuvant, PADRE, and His-tag were kept as in the N and C terminal end, and the LBL, MHC I, and MHC II were framed in different positions (11, 44) for the vaccine construction, as shown below, and the final constructed sequence was of 276 amino acids.

Furthermore, antigenicity and allergenicity revealed that the V2 combination was found to have the highest antigenic score among the different combinations, as shown in Supplementary Table 23 . Moreover, all the constructed vaccines in different forms have an antigenic nature and a non-allergenic feature, which ensures that the selected epitope is highly promising in various forms. These V2 combinations ( Figure 2 ) were similarly applied for the mutated vaccine formulation of 276 amino acids, and their antigenicity and allergenicity were analyzed ( Table 6 ). Furthermore, the physiochemical properties and solubility analysis revealed suitable properties of non-mutated ( Supplementary Table 24 ) and mutated vaccines, as in Table 6 .

For effectiveness, a potent vaccine must have a wide range of coverage (29). These eight epitopes (four MHC I and four MHC II) were examined together, and according to the restricted alleles, there was 97.41% coverage, which shows the broader coverage ( Figure 3 ) of the employed epitope in the vaccine formulation.

The secondary assessment revealed that the non-mutated vaccine has a helix, 23.91%; strand, 23.91%; and coil, 52.17% ( Supplementary Figure 5 ), whereas the mutated has a helix, 20.65%; strand, 25.72%; and coil, 53.62% ( Figure 4 ).

The model structure via Robetta (52) servers revealed a confidence score of 0.42 for the non-mutated and 0.41 for the mutated vaccine, which lies within the better-quality range. These models were further refined, and based on their various parameters, model 3 for the non-mutated ( Supplementary Figure 6A ) ( Supplementary Table 25 , highlighted in blue) and model 1 for the mutated vaccine ( Figure 5A ) ( Table 7 , highlighted in blue) were found suitable.

The structure quality validation via PROCHECK (54) demonstrated that the non-mutated vaccine has 87.3% residue in the most favored region, 8.6% residue in the additional allowed region, 1.8% residue in the generously allowed region, and 2.3% residue in the disallowed region ( Supplementary Figure 6B ), followed by 88.3% residue in the most favored region, 9.5% residue in the additional allowed region, 0.9% residue in the generously allowed region, and 1.4% residue in the disallowed region as in Figure 5B for the mutated vaccine. The Ramachandran plot shows that both non-mutated ( Supplementary Figure 6B ) and mutated ( Figure 5B ) vaccine models have only five and three residues in the disallowed regions and are scattered, suggesting less likely to cause significant structural instability. Moreover, most of the residue lies in the favored region, suggesting the overall reliable backbone geometry of the model (22, 68). Furthermore, the Z-score assessment done via ProSA-web (55) revealed that the non-mutated vaccine has a −6 score ( Supplementary Figure 6C ) and the mutated vaccine has a −5.54 score ( Figure 5C ); the negative score represents the superior structure model. Based on structural validation, the assessment demonstrated the good quality of the non-mutated and mutated vaccines (22).

The non-mutated and mutated vaccine structure was subjected to the Ellipro (56) server to compute the discontinuous epitope within the vaccine. The subjected non-mutated vaccine revealed that seven epitopes covered 139 amino acids; their range score varied from 0.618 to 0.815 ( Supplementary Table 26 ). In contrast, six epitopes were found for the mutated vaccine, covering 147 residues, followed by the score range from 0.588 to 0.967 ( Table 8 ). The discontinuous epitopes with both vaccines show that the construct vaccine will lead to a remarkable immune response (69).

The molecular activity of formulated non-mutated and mutated vaccines with the TLR3 was accomplished via ClusPro (7). The TLR3 can recognize double-stranded RNA (dsRNA) and single-stranded RNA (ssRNA) and is also vital in antiviral immune responses. Moreover, its activation stimulates dendritic cell activation mediated by HIV-1, which makes it an ideal target (7, 70). Among the generated multiple docked complexes of subjected TLR3 and vaccine, model 6 for the non-mutated ( Supplementary Table 27 ) and model 7 for the mutated vaccine ( Supplementary Table 28 ) were found most suitable, having high negative energies of −1,120.2 and −1,275.4 kcal/mol, respectively. The binding affinity of complexes was computed via PRODIGY (58), and the score was obtained at −12.8 kcal/mol (TLR3-Non-mutated) and −24.0 kcal/mol (TLR3-Mutated). These complexes were visualized for their various types of interaction followed by the H bond via PDBSum (59). The TLR3-Non-mutated complex shows 16 H bonds followed by 4 salt bridges and 196 non-bonded contacts as in Supplementary Figure 7 . In contrast, the TLR3-Mutated vaccine revealed 40 H bonds followed by 8 salt bridges and 364 non-bonded contacts, as in Figure 6 . Moreover, the interface residue is demonstrated in Supplementary Figure 7 , Figure 6 . The docking analysis revealed that the vaccine is strongly bound via molecular connection with TLR3, and the incorporated variability in the epitopes does not affect the interaction; rather, it improves, followed by a high number of hydrogen bonds.

The docked TLR3 with the non-mutated and mutated vaccines was analyzed via the Desmond software, followed by considering steps of the parameter (23, 61, 62) to examine their stability. The examination shows that the non-mutated and mutated vaccines remained bound with the TLR3 over the simulation period ( Figure 7 , Supplementary Figure 8 ). The RMSD investigation shows that the Cα of the mutated vaccine–TLR3 complex stabilized after 20 ns, followed by approximately 3.0–3.5 Å deviation, and the side chains were comparably slightly higher at approximately 4.5–5.0 Å, which shows the local conformational adjustments ( Figure 7A ), whereas the non-mutated vaccine–TLR3 complex was gradually stabilized after 20 ns and the Cα atoms rise between 6.0 and 6.5 Å, and the side changes merely followed a similar trend but are slightly higher and stabilized (6.5–7.0 Å) ( Supplementary Figure 8A ). The higher range of RMSD revealed great flexibility, and the complex maintained its structural stability (71, 72). Moreover, the RMSF investigation shows that the alpha of the mutated vaccine–TLR3 complex was less than 2 Å, and the side chain surpassed 4 Å at specific residues, which shows higher fluctuation ( Figure 7B ). In contrast, the alpha of the non-mutated vaccine–TLR3 complex remains below 3 Å, and their side chain was comparably higher with a minor exceeding 6–8 Å at certain regions ( Supplementary Figure 8B ). The minor high peaks in the RMSF of both docked complexes recommend confined rigidity, which is essential for interaction (60, 71, 72).

The ML accomplished vaccine immune activity and assisted the C-IMMsim server in considering the time steps of the injection interval, as in Figure 8 (Mutated) and Supplementary Figure 9 (Non-mutated). In the case of the non-mutated vaccine, the primary administration shows a high peak of antigen level (700,000 mL) and high generation of immunoglobin, followed by secondary and tertiary administration having an antigen count level of 500,000 each, which further instantly completely reduced, and further, the generated immunoglobins (IgM+IgG, IgM, IgG1+IgG2, IgG1, and IgG2) spiked (650,000) and continued to increase, as shown in Supplementary Figure 9A . In contrast, the mutated vaccine shows antigen counts of approximately 700,000, 300,000, and 50,000 per mL at the primary, secondary, and tertiary response levels, respectively. In contrast, the generated immunoglobin level shows a more promising spike (IgM+IgG, IgM, IgG1+IgG2, IgG1, and IgG2) followed by nearly 800,000, which is higher than the non-mutated immunoglobin level as in Figure 8A . Moreover, the generated cytokine and interleukins show the highest peaks (IFN-γ, IL-2, IL-4, and TNF-α) at nearly 450,000 ng/mL for non-mutated ( Supplementary Figure 9B ), nearly similar to the mutated vaccine ( Figure 8B ). The repeated exposure of the immunoglobin and cytokine level followed by steps of injection shows that the vaccine is capable of remarkable immune activity in both forms (Mutated, Figure 8 ; and Non-mutated, Supplementary Figure 9 ), and the incorporated variability does not reduce the vaccine’s effectiveness.

The queried non-mutated and mutated vaccine optimized sequence was 831 for each. The CAI value was 0.95 and GC% was 54.27 for the non-mutated vaccine. In contrast, for the mutated vaccine, the CAI was 0.95, and the GC% was 53.43, demonstrating the significant expression in the bacterial system of both vaccines as the obtained value lies in favor of the expression level. Furthermore, the optimized mutated and non-mutated vaccines (red) were cloned in the pET28a (+) vector in Figure 9 ; Supplementary Figure 10 .

Strain variability remains a significant challenge in HIV vaccine development. In this study, we successfully designed both non-mutated and mutated vaccine constructs, incorporating epitope variability to address this issue. The vaccines demonstrated remarkable immune activity, highlighting their potential effectiveness. Several steps of investigation and examination were employed via integrating the computational and immunoinformatic approach, which is associated with accuracy and promise. While the formulated vaccine revealed strong immune activity, future steps, including experimental validation, multi-strain efficacy, immune response evaluation, and clinical trials, are essential to ensure its protection and immune activity.