Protein Retrieval and Selection of Virulence Factors

Virulence-associated proteins were selected from the Virulence Factor Database (VFDB v.2025; ; accessed December 2025). The selected virulence factors included Streptococcal pyrogenic exotoxin B (SpeB), mitogenic factor (MF), streptolysin O (SLO), streptodornase (DNase), and NAD-glycohydrolase (NADase), selected based on their established functional relevance in pathogenicity. The corresponding amino acid sequences were retrieved in FASTA format from the NCBI Protein database (; accessed January 2026). A total of 100 non-redundant amino acid sequences were retrieved for each of the five major exoenzymes from the NCBI Reference Sequence (RefSeq) database (Release 229, January 2026). This sample size was chosen to balance computational feasibility with sufficient representation of sequence diversity for conservation analysis. All available full-length, experimentally verified sequences were initially collected, with partial, hypothetical, or ambiguous entries excluded. From this curated pool, redundancy was minimized using CD-HIT v.4.8.1 (; accessed January 2026) with a 90% identity cutoff, and 100 representative sequences per protein were selected to ensure a diverse yet manageable dataset. These sequences were subsequently used for multiple sequence alignment and epitope prediction analyses. The complete list of NCBI accession numbers for each protein is provided in the Supplementary Material.

Multiple Sequence Alignment and Conservation Analysis

Protein sequences corresponding to each Streptococcus pyogenes virulence factor were aligned using Clustal Omega v.1.2.4 (; accessed January 2026) to generate multiple sequence alignments (MSA). Sequence conservation was analyzed using the Protein Variability Server (PVS) (; accessed January 2026), where positional variability was quantified using the Shannon entropy ( H ) index. A fixed cutoff of H ≤ 0.1 was applied to define highly conserved residues, consistent with established thresholds in previous studies.13,14 Variable positions exceeding this threshold were masked and excluded, ensuring that only strongly conserved fragments (≥9 amino acids in length) were advanced for downstream MHC binding and immunogenicity analyses. The corresponding Shannon entropy profiles for each protein are provided in the Supplementary Material.

Epitope Prediction, MHC Binding Evaluation, and PTM Screening

Conserved regions from the selected virulence proteins were submitted to the Immune Epitope Database (IEDB) v2.28 (; accessed January 2026) for cytotoxic T-lymphocyte (CTL) epitope prediction. Predictions were performed using the NetMHCcons algorithm with default parameters, which integrates NetMHCpan v3.0 to assess MHC class I binding. Predicted epitopes were ranked based on combined scores incorporating MHC binding affinity, proteasomal cleavage likelihood, and TAP transport efficiency. A total of 36 globally and regionally common HLA class I alleles (HLA-A and HLA-B) were included to ensure broad population coverage. Predicted epitopes were filtered according to the following criteria: (1) MHC binding affinity IC₅₀ ≤ 500 nM, (2) proteasome cleavage score ≥ 1.0, (3) TAP transport score ≥ 1.0, and (4) only non-redundant sequences were retained. For reference, NetMHCpan output scores typically range from 0 to 3.0, with higher values indicating stronger predicted immunogenic potential. Candidate peptides exceeding these thresholds were further screened for antigenicity, allergenicity, toxicity, and cross-reactivity using the IEDB analysis pipeline.

Because epitope-binding predictions are generated by machine learning models rather than conventional hypothesis-driven statistical tests, standard multiple-comparison corrections (e.g., FDR) are not directly applicable. To reduce false positives, only epitopes consistently identified as strong binders (IC₅₀ ≤ 500 nM) by two independent algorithms (NetMHCpan and NetCTL) were retained for downstream analysis. This concordance-based approach is commonly used in immunoinformatics studies and is supported by empirical evidence showing that epitopes predicted by multiple independent tools have a higher likelihood of representing true MHC binders.15

To further ensure structural stability, all candidate epitopes were assessed for post-translational modifications (PTMs) using UniProt v.2025_02 (). Potential PTM sites within the selected virulence proteins were examined to ensure that predicted epitopes did not overlap with modified residues that could interfere with antigen processing or MHC binding. PTM annotations were retrieved from the UniProtKB database (release 2025_01; accessed January 2026) using the “Modified residue” and “Glycosylation” feature types. Cross-verification using dbPTM version 2024 () confirmed the absence of annotated or predicted PTMs within the final epitope regions, ensuring structural and functional integrity of the selected peptides. Any epitope overlapping with modified residues was excluded to avoid potential processing interference.

Evaluation of Allergenicity, Toxicity, Cross-Reactivity, and Population Coverage

All predicted CTL epitopes were screened for allergenicity and toxicity. AllerTOP v.2.1 () and AllergenFP v.1.1 () apply auto-cross covariance (ACC) transformation and E-descriptor–based encoding of amino acids (hydrophobicity, size, and secondary-structure propensities) followed by machine-learning classification— k-nearest neighbors (AllerTOP) or Tanimoto similarity (AllergenFP)—using a default probability cutoff of 0.5. Toxicity was predicted using ToxinPred2 (), an SVM-based model operating at its default confidence threshold. These tools exhibit an average false-positive rate of 10–15%. Only epitopes predicted as non-allergenic and non-toxic by all tools were retained for downstream analyses.

To ensure non-reactivity with human proteins, each epitope was submitted to BLASTp analysis against the Homo sapiens (TaxID: 9606) proteome from the NCBI RefSeq protein database (Release 229, March 2025) using the NCBI BLAST tool (; accessed May 2025). Epitopes with E-values less than 1.0 were considered potentially cross-reactive and were excluded. Only those with E-values greater than 1.0 were retained as safe and non-homologous to human proteins.

The shortlisted epitopes were evaluated using the IEDB Population Coverage Tool (; accessed January 2026) to estimate their representation across HLA alleles in the Philippines, Southeast Asia, and global populations. This analysis provides point estimates of predicted population coverage, reflecting the computationally inferred potential immunogenicity of the selected epitopes across diverse genetic backgrounds. Confidence intervals were not calculated, as the current version of the IEDB tool does not support bootstrapping or uncertainty modeling. Consequently, the reported percentages should be interpreted as approximate estimates, which may vary with updates to allele frequency data or inclusion of additional population datasets.

Molecular Docking Analysis of Epitope-MHC Complexes

The top five epitopes exhibiting the highest predicted global population coverage were selected for molecular docking with their respective MHC class I alleles. Docking was performed using GalaxyPepDock (build 2023_05_29; ; accessed January 2026), a template-based peptide–protein docking tool. The three-dimensional structures of the corresponding HLA molecules were obtained from the Protein Data Bank (; accessed January 2026) and preprocessed in UCSF ChimeraX v.1.10, where only Chain A, constituting the peptide-binding domain, was retained. Non-relevant chains, ligands, and water molecules were removed prior to docking. All GalaxyPepDock runs used the server’s default parameters, which include automatic template selection from a database of experimentally determined peptide–protein complexes, rigid-body docking guided by interaction similarity, and flexible energy-based refinement of peptide and side-chain conformations. The server generates up to 10 candidate complex models per docking run, ranked by energy and interaction scores. Re-docking validation of native peptide–protein complexes was performed to confirm the reliability of the docking protocol, and all top-ranked poses were analyzed for structural consistency.

GalaxyPepDock employs a template-based deterministic docking algorithm, matching input sequences to structurally similar complexes rather than relying on stochastic sampling; therefore, random seeding is not applicable. All docking runs were conducted under uniform default parameters to ensure reproducibility and comparability of results. The resulting complexes were visually validated to confirm epitope localization within the peptide-binding groove, flanked by the characteristic α-helical regions. For each peptide–HLA pair, the top 10 docking models were retained to compute mean ± SD of hydrogen bond statistics and evaluate interfacial interaction stability.

To evaluate the reliability of the docking protocol, a re-docking validation was conducted using the crystallized peptide–HLA-A*02:01 complex (PDB ID: 1I4F, 1.4 Å resolution). The native peptide was removed from the structure, and the receptor was used as the docking target. Docking was performed in GalaxyPepDock under parameters identical to those in the main analysis. The resulting docked pose was superimposed onto the crystal structure using receptor Cα atoms, and the root-mean-square deviation (RMSD) between the docked and native peptide backbones was computed using PyMOL and UCSF ChimeraX. The re-docking yielded a Cα RMSD of 1.78 Å (all-atom RMSD = 2.31 Å), indicating excellent reproduction of the experimental binding conformation and confirming the reliability of the GalaxyPepDock protocol.

To evaluate the thermodynamic stability of the docked complexes, binding affinity was estimated using the PRODIGY server version 2.3.0 (; accessed January 2026). The Gibbs free energy of binding (ΔG) and the dissociation constant (K) at 37 °C were computed to quantify interaction strength. Complexes exhibiting lower ΔG values and K D < 5 × 10⁻⁷ M were considered to possess strong and stable binding, indicative of favorable predicted immunogenic potential and suitability for vaccine development. The PRODIGY algorithm estimates ΔG based on interfacial contacts and solvent-accessible surface area, with an associated root-mean-square error (RMSE) of approximately ± 1.9 kcal/mol.

Design of the CTL-Based Multi-Epitope Vaccine Construct

A multi-epitope vaccine candidate (SpMEV) was designed by assembling computationally identified cytotoxic T lymphocyte (CTL) epitopes originating from different virulence-associated exoenzyme proteins of . To preserve the immunogenicity of individual epitopes and minimize junctional immunogenicity, the CTL epitopes were concatenated using AAY linkers, which also facilitate effective antigen processing. To boost the overall immunostimulatory potential of the construct, human β-defensin 3 (hBD-3) was incorporated as an adjuvant. hBD-3 is a host defense peptide that functions as a Toll-like receptor (TLR) agonist and possesses inherent antimicrobial as well as immune-enhancing properties. The full-length amino acid sequence of hBD-3 was fused at both the N- and C-termini of the construct via a rigid EAAAK linker, ensuring structural stability and functional independence between the adjuvant and epitope array. A valine residue was added at the extreme N-terminus to enhance proteolytic resistance and prolong protein half-life within the host cell. For downstream purification applications, a C-terminal 6×His affinity tag was appended.

Safety Profiling: Allergenicity, Antigenicity, and Toxicity Assessment

The complete amino acid sequence of SpMEV was subjected to computational evaluation for antigenicity, allergenicity, and toxicity using established web-based prediction tools. Candidate suitability for further development was determined on the basis of being antigenic, non-allergenic, and non-toxic according to the default threshold scores provided by each algorithm.

Prediction of Physicochemical Characteristics

Key physicochemical descriptors of SpMEV, including molecular weight, theoretical isoelectric point (pI), amino acid composition, estimated half-life, instability index, aliphatic index, and the grand average of hydropathicity (GRAVY), were calculated using the ProtParam tool on the ExPASy server v.3.0 (; accessed January 2026).

Structural Characterization: Secondary and Tertiary Structure Prediction

Secondary structure profiling of SpMEV was performed using PSIPRED v4.0 (; accessed June 2025), providing the predicted distribution of α-helices, β-strands, and random coils. PrDOS (; accessed January 2026) was employed to identify intrinsically disordered regions, while solvent accessibility of amino acid residues was analyzed using NetSurfP-3.0 (; accessed January 2026).

Three-dimensional structure modeling was conducted via a multi-template strategy. Initial structural models were generated using I-TASSER (; accessed January 2026) and GALAXY-TMB (; accessed January 2026), followed by refinement with GalaxyRefine2 (; accessed January 2026). Model validation was carried out through multiple approaches: (1) ERRAT (; accessed January 2026) to examine non-bonded atomic interactions; (2) ProSA-web (; accessed January 2026) for overall Z-score evaluation against crystallographic/NMR structures; and (3) PROCHECK (; accessed January 2026) to assess stereochemical quality via Ramachandran plot statistics.

The final validated 3D structure of SpMEV was visualized in UCSF ChimeraX v.1.10 (), where the structural organization and surface accessibility of predicted CTL epitopes were examined to ensure adequate presentation for immune recognition.

Identification and Immunogenic Profiling of Conserved Cytotoxic Epitopes from Exoenzymes

To ensure broad protection across diverse strains, conserved antigenic regions were identified for each target protein using the Protein Variability Server, retaining fragments with a Shannon entropy ≤ 0.1 and a minimum length of nine amino acids. Low entropy indicates strong evolutionary conservation, and a nine-residue length corresponds to the optimal size for MHC class I binding. The ≤ 0.1 cutoff reflects minimal sequence variability and is widely used in immunoinformatics to define highly conserved regions suitable for vaccine targets. This approach prioritizes functionally constrained regions with high cross-strain immunogenic potential, thereby reducing the risk of immune escape in bacterial pathogens 13,14.

Predicted CTL epitopes from these conserved regions were screened for MHC class I affinity using NetMHCpan within the IEDB. Peptides with an IC₅₀ ≤ 500 nM were classified as strong binders, and only those with favorable proteasomal cleavage and TAP transport scores (≥ 1.0) were retained, ensuring their likelihood of natural processing and presentation 20,21. To maintain biological relevance, predicted epitopes were cross-checked with post-translational modification (PTM) sites from UniProt, as PTMs can alter peptide stability and HLA binding 22. Epitopes overlapping PTM sites were excluded, particularly for SpeB, MF, and SLO.

Safety assessment included filtering for allergenicity, toxicity, and human cross-reactivity to minimize the risk of off-target immune responses. Allergenicity was assessed via AllerTOP v.2.1 and AllergenFP v.1.1, and peptides flagged as probable allergens by either tool were excluded to avoid IgE-mediated hypersensitivity responses, commonly elicited by small antigens capable of cross-linking IgE on mast cells 23. ToxinPred2 was used to assess peptide toxicity based on amino acid composition and motifs, retaining only non-toxic sequences to minimize off-target or cytotoxic effects 24. To prevent autoimmune risk, all non-toxic epitopes were screened against the human proteome using BLASTp, applying a stringent E-value cut-off (>1) to eliminate sequences with significant similarity to endogenous human proteins. Peptides exhibiting E-values less than 1 were flagged as potentially cross-reactive and excluded, as such homology may trigger autoreactive T-cell responses and increase the risk of autoimmune complications 25. The retained epitopes exhibited E-values ranging from 1.2 to 73, confirming the absence of relevant homology and minimizing the potential for cross-reactive autoimmunity.

Among the five exoenzymes, SLO yielded the highest number of conserved and immunogenic CTL epitopes (n = 35), underscoring its evolutionary stability and immunodominance across >98% of isolates 26. SpeB (n = 19) and DNase (n = 16) contributed epitopes consistent with their established roles in immune evasion via proteolysis and NET degradation, respectively. NADase (n = 14) and MF (n = 10) provided complementary epitopes associated with intracellular survival and superantigenic activity. The final list of epitopes per virulence factor is summarized in Table 1.

Evaluation of HLA Binding Promiscuity and Population Coverage

In this study, the final CTL epitopes were analyzed using the IEDB Population Coverage Tool, which integrates experimentally verified HLA frequency data across diverse ethnic groups. Results revealed that SLO-derived epitopes achieved the highest global coverage (91.07%), followed by SpeB (78.00%), DNase (77.02%), MF (71.72%), and NADase (70.39%) (Table 2). These values represent the proportion of the global population predicted to mount a CTL response to at least one epitope from each antigenic source.

Regional analysis further revealed allele-specific variations influencing local immunogenic potential. Using IEDB-reported HLA class I frequencies (11 documented alleles for the Philippines), DNase and MF epitopes showed limited combined coverage (13.51%), reflecting regional allele bias toward non-dominant supertypes 27. In contrast, SLO and NADase epitopes retained high compatibility (69.86%) with locally prevalent alleles, suggesting favorable regional immunogenicity.

Structural Binding Validation via Molecular Docking

Molecular docking was employed to evaluate the structural compatibility and binding strength of the top five CTL epitopes from exoenzymes (SpeB, MF, SLO, DNase, and NADase) with their respective HLA class I alleles. These epitopes were pre-screened based on high MHC affinity (IC₅₀ ≤ 500 nM), non-allergenicity, non-toxicity, and HLA-binding promiscuity, ensuring immunologically robust candidates for downstream validation.

All epitopes were found to occupy the canonical MHC class I peptide-binding cleft, forming stabilizing hydrophobic and hydrogen bonds consistent with known HLA–peptide interactions (Figure 1). Anchor residues located at positions 2, 5, and the C-terminus mediated the strongest contacts, reinforcing peptide stability and proper groove alignment 28. No structural clashes or misorientations were observed, and all peptides adopted TCR-accessible conformations, supporting their potential for effective antigen presentation and CD8⁺ T-cell activation 29.

To assess thermodynamic favorability, binding free energy (ΔG) and dissociation constants (K) were calculated using the PRODIGY server. All epitope–HLA complexes exhibited negative ΔG values (≤ –8.6 kcal/mol) and submicromolar K values (≤ 880 nM), indicating spontaneous and stable binding under physiological conditions (Table 3). The NADase–HLA-B15:01–TQYTESMVY complex displayed the strongest interaction (ΔG = –12.8 kcal/mol; K = 0.99 nM), while the DNase–HLA-B15:01–KMIDMSAGY complex exhibited the weakest yet favorable binding (ΔG = –8.6 kcal/mol; K = 880 nM). These values fall within the range expected for stable protein–peptide complexes, validating the docking accuracy and biological relevance of the modeled interactions.

A strong inverse correlation was observed between the predicted Gibbs free energy of binding (ΔG) and the dissociation constant (K) across all epitope–HLA complexes (Figure 2). Complexes with more negative ΔG values exhibited lower K values, reflecting higher binding affinity and complex stability. Among the analyzed virulence factors, epitopes derived from NADase, SLO, and SpeB demonstrated the strongest binding interactions (ΔG ≤ –11 kcal/mol; K in the nanomolar range), supporting their immunogenic potential.

Hydrogen Bond Geometry and Stability of HLA–Epitope Complexes

The geometric parameters of hydrogen bonds provide critical insights into their strength and role in molecular recognition. Empirical studies have established that donor–acceptor (D···A) distances of 2.7–3.3 Å typically denote stable hydrogen bonds, while distances below 3.0 Å indicate stronger and more favorable interactions 30. Likewise, hydrogen–acceptor (H···A) distances of 1.6–2.5 Å are characteristic of stable hydrogen bonding, with shorter values in the 1.6–2.2 Å range representing near-optimal interactions when complemented by favorable bond angles. These geometric parameters are particularly significant in peptide–MHC (pMHC) systems, where robust hydrogen bonding helps anchor conserved peptide residues within the MHC binding groove, promoting specificity and long-term complex stability 30.

As summarized in Table 4, all modeled complexes displayed extensive hydrogen bonding networks, ranging from 10 to 30 hydrogen bonds per complex. Donor–acceptor distances spanned approximately 2.70–4.92 Å, and hydrogen–acceptor distances ranged from 1.84–3.95 Å. The lower-bound values fall within the strong hydrogen bond regime, especially those near 1.84–2.0 Å (H···A) and 2.7–3.0 Å (D···A), suggesting favorable enthalpic contributions to binding affinity. Notably, DNase-derived epitopes in complexes C2 and C3 exhibited the highest number of hydrogen bonds (28 and 30, respectively), indicating highly stabilized interactions with their respective HLA alleles. Such dense hydrogen bond networks are known to increase complex half-life and enhance immunogenicity in pMHC systems, as they promote stable epitope anchoring and efficient presentation to cytotoxic T lymphocytes 27. Conversely, interactions near the upper distance limits (up to ~3.9 Å for H···A and ~4.9 Å for D···A) likely correspond to weaker or water-mediated hydrogen bonds. While individually less stabilizing, these interactions contribute to the overall orientation, flexibility, and conformational adaptability necessary for effective T-cell receptor recognition.

SpMEV Construct Design and Structural Features

A multi-epitope peptide vaccine candidate against Streptococcus pyogenes (SpMEV) was developed by systematically integrating six cytotoxic T-lymphocyte (CTL) epitopes that exhibited the most favorable thermodynamic characteristics, specifically the lowest predicted Gibbs free energy (ΔG) and dissociation constant (K) values. These epitopes, derived from immunodominant exoenzymes, were arranged in tandem and connected by AAY linkers 31. The AAY linker facilitates proteasomal cleavage and efficient presentation via the MHC class I pathway 32, ensuring proper processing of each epitope while reducing the formation of junctional epitopes that could compromise immunogenicity. While this design minimizes the formation of junctional epitopes, it is important to note that novel junctional sequences may still arise and influence epitope processing or immunodominance, warranting future experimental validation. The construct was deliberately designed as a CTL-focused vaccine to stimulate robust cell-mediated immunity against , particularly targeting the intracellular persistence phase of infection 33.

To enhance immunostimulatory potential, human β-defensin-3 (hBD-3) was incorporated at both the N- and C-termini of the construct as an intrinsic adjuvant. hBD-3 was selected due to its documented antimicrobial activity and its capacity to promote antigen presentation and dendritic cell maturation 34. Previous studies have reported hBD-3–mediated maturation of Langerhans-like dendritic cells and adjuvant effects in cutaneous vaccination models, whereas other findings indicate context-dependent immunosuppression, modulation of innate signaling pathways, and potential pro-inflammatory responses 35,36.

Rigid EAAAK linkers were employed to join the hBD-3 adjuvant domains to the CTL epitope cluster, ensuring structural independence, optimal spatial separation, and reduced interference between the epitopes and adjuvant segments. A C-terminal valine residue was appended to the construct to improve structural stability and protease resistance, extending the potential half-life of the construct. In addition, a polyhistidine (6×His) tag was introduced to facilitate downstream purification via affinity chromatography, improving the construct’s feasibility for recombinant expression and large-scale production. The schematic diagram of the final SpMEV construct (Figure 3) consisted of 170 amino acids organized in a rational architecture balancing immunogenicity, stability, and manufacturability.

Characterization of Immunogenic and Safety Profiles of the SpMEV Construct

The designed multi-epitope vaccine (SpMEV) construct was subjected to immunoinformatic evaluation to assess its antigenicity, allergenicity, toxicity, and cross-reactivity. Antigenicity analysis using VaxiJen yielded a score of 0.5387, which surpasses the default threshold of 0.4, while ANTIGENpro predicted a significantly higher value of 0.9137. These results suggest that the construct possesses strong intrinsic immunogenic potential, an essential feature for eliciting protective immune responses 23.

Allergenicity prediction indicated that the construct is a probable non-allergen, supported by a Tanimoto coefficient of 0.4368 when compared to the most similar human protein, the cystic fibrosis transmembrane conductance regulator (CFTR). This relatively low similarity index supports its classification as non-allergenic, reducing the likelihood of hypersensitivity reactions. Importantly, AllerCatPro also classified the construct as a probable non-allergen, with no significant sequence similarity detected at the stringent E-value threshold of 0.001, further strengthening confidence in its safety profile.

Further evaluation using ToxinPred2 classified the construct as non-toxic, strengthening its safety profile for potential applications. Cross-reactivity screening also confirmed that the construct does not share significant similarity with human proteins, thereby minimizing the risk of autoimmune responses. Collectively, these results indicate that the SpMEV construct is predicted to be highly antigenic, non-allergenic, non-toxic, and non-cross-reactive, supporting its suitability as a safe and effective vaccine candidate. Structural validation (secondary and tertiary) was subsequently conducted to ensure the stability and reliability of the construct.

Physicochemical Properties of the SpMEV Construct

The final SpMEV construct comprises 170 amino acids, with a theoretical molecular weight of 19.37 kDa and an isoelectric point (pI) of 9.89, suggesting it is a basic and predominantly cationic protein at physiological pH. The overall amino acid composition indicates enrichment in positively charged residues, contributing to a net charge favorable for electrostatic interactions with antigen-presenting cells. The predicted extinction coefficient (26,080 M⁻¹ cm⁻¹) and instability index (39.97) indicate good structural stability, while an aliphatic index of 65.59 and a GRAVY score of –0.47 suggest moderate thermostability and hydrophilicity, respectively—properties generally consistent with soluble and immunogenic recombinant proteins. The estimated half-lives across mammalian, yeast, and bacterial systems (≥ 10 h) further support the construct’s potential suitability for heterologous expression. The detailed ProtParam outputs (amino acid composition, charge distribution, extinction coefficient assumptions, and atomic composition) are provided in the supplementary material. These summarized physicochemical characteristics collectively indicate that the SpMEV construct is predicted to be a stable, soluble, and moderately thermostable vaccine candidate suitable for downstream expression and validation.

Secondary Structure of the SpMEV Construct

The secondary structure profile of the designed SpMEV vaccine revealed a favorable balance of α-helices, β-strands, and coil regions (Figure 4). From an immunological perspective, such structural diversity is advantageous because it provides both rigidity and flexibility within the vaccine construct. The α-helical and β-strand elements contribute to the stability of the overall fold, ensuring that epitopes are maintained in a defined conformation that can be consistently recognized by the immune system. In contrast, coil regions provide structural adaptability, often facilitating the surface presentation of epitopes to major histocompatibility complex (MHC) molecules. This combination of ordered and flexible domains reflects a rational design principle in epitope-based vaccines, where stability and accessibility must be carefully balanced 37.

Disorder analysis indicated that most residues adopt structured conformations with only localized flexibility near the terminal polyhistidine tag. Such intrinsic disorder is not detrimental; in fact, moderate disorder in vaccine constructs can enhance antigen processing by the proteasome and promote efficient MHC presentation. This aligns with immunological evidence that disordered or flexible peptide regions are more prone to proteolytic cleavage, thereby increasing the likelihood of epitope liberation and presentation to cytotoxic T lymphocytes (CTLs) 38.

Surface accessibility predictions further supported the immunogenic potential of the construct, as the majority of CTL epitopes were located in solvent-exposed regions. Epitope exposure at the protein surface is essential for recognition by antigen-presenting cells and subsequent T-cell priming 20. Meanwhile, the burial of hydrophobic residues within the protein core suggests proper folding and minimized aggregation propensity, which are critical for vaccine stability and manufacturability.

Tertiary Structure of the SpMEV Construct

The quality of the predicted three-dimensional protein structure was assessed before and after refinement using stereochemical and energetic validation tools (Figure 5). The initial Ramachandran plot analysis revealed that 92.9% of residues were located in favored regions, with a minor fraction occupying disallowed conformational space. Following successive refinements using ModRefiner from the I-TASSER platform and GalaxyRefine2 from the GalaxyWeb server, the stereochemical accuracy improved substantially, yielding 98.7% of residues within the favored regions and, notably, a complete absence of residues in disallowed regions. This marked enhancement reflects the successful correction of backbone torsional strain and indicates a more realistic accommodation of secondary structure elements within energetically permissible conformations 39.

Consistent with these improvements, the ERRAT quality factor increased from 82.69% in the unrefined model to 98.60% after refinement. Since ERRAT evaluates the statistics of non-bonded interactions, this increase signifies a substantial reduction in steric clashes and non-ideal contacts, leading to improved atomic packing and greater structural stability 40. Such a refinement outcome is characteristic of the effective optimization of hydrogen bonding networks and hydrophobic core interactions, both of which are crucial for stabilizing protein folds.

The improvement was further supported by ProSA analysis, where the z-score shifted from -3.87 to -5.23. As ProSA z-scores benchmark protein models against experimentally resolved structures of comparable size, the refined structure’s more negative value suggests a closer approximation to a native-like fold with improved thermodynamic plausibility 41. Importantly, the convergence of these independent validation metrics strongly indicates that the applied refinement protocols were not only effective in correcting local stereochemical errors but also in enhancing the global energetics of the protein model.

The combined use of ModRefiner and GalaxyRefine2 likely underpins the observed improvements, as each tool contributes distinct refinements: ModRefiner emphasizes atomic-level geometry and torsional accuracy, while GalaxyRefine2 performs iterative side-chain repacking and relaxation of the entire structure 42. Consequently, these refinements yielded a model with structural quality metrics approaching those of experimentally determined proteins. The resulting structure, free from disallowed torsions and exhibiting high-quality packing statistics, provides a more reliable basis for downstream functional annotation, docking simulations, or dynamic studies. The refined three-dimensional construct of the designed SpMEV with labels is shown in Figure 6.

Determinants of Epitope Selection and Immunological Relevance

Notably, the number of predicted CTL epitopes varied among the selected exoenzymes, with streptolysin O (SLO) yielding a higher epitope count compared to the other proteins. While a greater epitope yield may contribute to broader HLA population coverage, epitope abundance alone does not necessarily correlate with immunodominance or immunogenicity. Antigen processing efficiency, proteasomal cleavage, TAP transport, MHC class I binding stability, and T-cell receptor recognition collectively influence the immunological relevance of individual epitopes. Therefore, the differential epitope distribution observed across exoenzymes should be interpreted as a predictive indicator rather than a definitive measure of vaccine efficacy.

The cumulative filtering process substantially reduced the dataset but improved its overall quality, yielding epitopes that are non-allergenic, non-toxic, and non-cross-reactive. This stringent multi-tiered screening ensures that only biologically safe and immunologically relevant peptides are advanced for downstream analyses, thereby enhancing their translational feasibility for vaccine development. Moreover, this rigorous selection framework strengthens the specificity, safety, and immunological fidelity of the vaccine construct, minimizing the risk of adverse immune reactions and increasing its potential for regulatory approval and clinical success.

Implications of HLA Binding Promiscuity and Population Coverage in Multi-Epitope Vaccine Design

A central challenge in epitope-based vaccine design is overcoming the extensive polymorphism of HLA class I molecules, which differ substantially across global populations.26 Because cytotoxic T lymphocytes (CTLs) recognize antigens only when presented by compatible HLA alleles, epitope selection must prioritize promiscuous peptides capable of binding multiple alleles to achieve broad immunogenic coverage. Previous analyses indicate that targeting a subset of such promiscuous epitopes can yield over 90% global population coverage, although additional allele representation remains critical in genetically diverse regions such as Africa and Southeast Asia.43

Compared with recent multi-epitope vaccine designs against 43, the present work introduces three notable advances. First, the dominant inclusion of SLO-derived epitopes confers markedly higher global coverage (>90%) than earlier M-protein-focused constructs. Second, the downstream incorporation of β-defensin-3 as an intrinsic adjuvant is designed to enhance both mucosal and cellular immune activation, addressing limitations in prior models that lacked endogenous adjuvanticity. Third, the integration of regional HLA allele frequency analysis, particularly highlighting Philippine-prevalent alleles (HLA-A*11:01, HLA-B*15:02), provides a geographically nuanced understanding often absent in global GAS vaccine pipelines.

The strong global population coverage of SLO epitopes underscores their value as universal vaccine components, consistent with their high conservation and immunodominance as cholesterol-dependent cytolysins eliciting potent CD8⁺ T-cell responses.2 Similarly, SpeB and DNase epitopes demonstrated broad multi-allelic binding capacity (78.00% and 77.02%), reflecting their conserved and surface-accessible nature, which favors efficient MHC class I presentation.17 Although MF and NADase achieved slightly lower coverage (71.72% and 70.39%), both exceed the 70% global feasibility threshold.44 MF's modest coverage likely reflects its lower epitope yield, whereas NADase contributes epitopes essential for intracellular immune modulation and synergistic function with SLO.19 These findings demonstrate that incorporating SLO as the dominant epitope source, alongside complementary antigens such as SpeB, DNase, MF, and NADase, yields a multi-antigen, multi-epitope framework capable of achieving both global and region-specific population coverage.45

Structural and Thermodynamic Validation of Epitope–HLA Interactions

Molecular docking provides structural and energetic validation of peptide–HLA interactions, enabling prediction of binding affinity, orientation, and stability.46 While docking offers valuable insights into epitope–HLA binding behavior, it represents a static snapshot of molecular interactions. To fully assess the stability, flexibility, and dynamic behavior of these complexes under physiological conditions, molecular dynamics simulations would be necessary in future studies.

Thermodynamic principles further corroborate the docking results, as spontaneous binding occurs when ΔG < 0, with more negative values indicating stronger association. Correspondingly, the low K values observed reflect tight ligand–receptor binding and minimal complex dissociation, hallmarks of immunogenic stability. The consistency between ΔG, K, and docking conformations confirms that the selected epitopes form energetically favorable, structurally stable, and TCR-compatible complexes with their HLA alleles.

Hence, SLO-, SpeB-, and NADase-derived epitopes demonstrated the most stable and energetically favorable docking profiles, aligning with their previously observed high immunodominance and population coverage. It is important to note that the ΔG values predicted by the PRODIGY server have an estimated uncertainty of ±1.9 kcal/mol (RMSE), reflecting inherent model variability. Consequently, small differences in docking scores between epitopes should be interpreted cautiously, as they may fall within the prediction error range rather than indicating true differences in binding strength. Nonetheless, the overall trend of consistently low ΔG and K values across the selected complexes supports their strong binding potential and biological relevance. Collectively, these structural and thermodynamic validations confirm that the selected epitopes possess robust MHC class I binding, high structural fidelity, and strong immunogenic potential, underscoring their suitability for rational design of a multi-epitope vaccine against .

Design Rationale, Safety Considerations, and Translational Implications of the SpMEV Construct

While humoral coverage through helper T-cell (HTL) and B-cell epitopes is important for achieving broad protection, this study specifically optimized the cytotoxic arm, with inclusion of HTL/B-cell epitopes planned in future construct iterations. The deliberate focus on CTL activation reflects the need to target intracellular persistence mechanisms of , where cytotoxic immunity plays a central role in pathogen clearance.

Given the dual immunomodulatory nature of human β-defensin-3, several precautions were adopted to mitigate potential autoreactive or adverse effects. The full-length human hBD-3 sequence was employed to avoid neo-epitope formation, and all vaccine peptides were rigorously screened against the human proteome to exclude potential autoreactive epitopes. Additionally, computational docking and epitope mapping were performed with the understanding that predictions cannot fully capture immunogenic variability, cytokine milieu effects, or host-specific responses, which may influence the overall adjuvant activity.

Future experimental validation should include comprehensive safety and efficacy assessments. These may encompass dose-escalation studies in murine models, profiling of Th1/Th2/Th17/Treg cytokines, autoantibody screening, histopathological examinations, and clinical monitoring for autoimmune manifestations. Importantly, these studies would clarify whether hBD-3 functions reliably as a safe endogenous adjuvant across different tissues and hosts or if alternative adjuvants with more predictable immunostimulatory profiles should be considered. Until such empirical evidence is available, any translational claims regarding hBD-3 use in humans should be interpreted with caution.

Immunogenicity and Safety Implications of the SpMEV Profile

The evaluation of antigenicity, allergenicity, toxicity, and cross-reactivity provides essential preliminary validation of the SpMEV construct prior to experimental testing. The high antigenicity scores obtained from both VaxiJen and ANTIGENpro indicate strong intrinsic immunogenic potential, which is a critical requirement for effective vaccine design. These findings support the rationale that the selected epitope combination can elicit a robust immune response upon administration.

The absence of predicted allergenicity and toxicity further reinforces the biosafety of the construct. The low Tanimoto similarity index and negative AllerCatPro prediction suggest minimal risk of hypersensitivity or off-target immune activation, which is a common limitation in peptide-based vaccines. Additionally, the lack of significant similarity with human proteins reduces the likelihood of autoimmune cross-reactivity, an important consideration for translational vaccine development.

Together, these results demonstrate that the SpMEV construct satisfies key immunological and safety benchmarks required for further development. While computational predictions provide valuable insight, experimental validation remains essential to confirm these findings under physiological conditions. Nonetheless, the favorable profile supports progression of the SpMEV construct toward structural validation, immunogenicity assays, and subsequent evaluation.

Structural Features and Implications for Vaccine Design

The secondary structure composition, limited intrinsic disorder, and favorable solvent accessibility profile indicate that the SpMEV vaccine construct is structurally stable yet sufficiently flexible to allow efficient epitope processing and presentation. These features are consistent with effective immunogen design principles, supporting the construct's potential to elicit robust and targeted immune responses.20,38

The balance between structural rigidity and flexibility is particularly important in multi-epitope vaccines, where excessive rigidity can hinder antigen processing, while excessive disorder may compromise structural integrity. In this construct, the predominance of ordered secondary structural elements ensures conformational stability, while localized disordered regions may facilitate proteasomal cleavage and MHC loading. This structural configuration aligns well with current understanding of antigen processing pathways and reinforces the suitability of the SpMEV construct for further experimental validation.

The observed improvements in stereochemical accuracy, atomic packing, and global energetics strongly support the structural reliability of the SpMEV construct. The increase in residues within favored Ramachandran regions, elevated ERRAT quality factor, and more negative ProSA z-score collectively indicate a model approaching native-like conformations. These structural enhancements are critical for multi-epitope vaccine design, as accurate three-dimensional folding ensures proper epitope orientation, stability, and accessibility for immune recognition.39

The refinement protocols applied—ModRefiner for atomic-level geometry and GalaxyRefine2 for side-chain repacking and relaxation—demonstrate the importance of combining complementary computational approaches to achieve high-quality protein models. A structurally robust tertiary fold is essential not only for stability during storage and delivery but also for reliable downstream applications such as molecular docking, dynamic simulations, and experimental expression and .

Although these findings are based on computational predictions, the refined secondary and tertiary structures provide a strong foundation for further experimental validation, including molecular dynamics simulations, recombinant expression, and functional immunogenicity assessments. Ultimately, these high-quality structural features support the translational potential of the SpMEV construct as a rationally designed vaccine candidate.