Immunoinformatics-guided design and in silico evaluation of BAFF- and APRIL-adjuvanted multi-epitope vaccine candidates against human metapneumovirus

1 Introduction

Human metapneumovirus (HMPV) is a leading cause of acute respiratory tract infections across all age groups and is particularly associated with severe disease in infants, older adults, and immunocompromised individuals (1). Since its first identification in 2001, HMPV has been recognized as a ubiquitous pathogen with substantial global health and economic impacts (2, 3). In pediatric populations, HMPV accounts for approximately 4-13% of hospitalizations associated with lower respiratory tract infections (LRTIs) in children under five years of age. HMPV is currently recognized as one of the leading viral causes of pediatric LRTIs and represents the second most common etiological agent after RSV, contributing substantially to hospitalization rates and disease severity. Globally, HMPV was estimated to cause approximately 11 million LRTI episodes in children younger than five years, resulting in nearly 502,000 hospitalizations and 11,300 deaths worldwide. These findings highlight the significant clinical and public health burden of HMPV, positioning its global impact between that of hRSV and influenza (1).

HMPV plays a critical role in modulating the host immune response by interfering with antiviral signaling pathways, suppressing interferon production, and impairing the function of immune cells (1). These immune evasion strategies contribute to the relatively weak and short-lived protective immunity observed following natural HMPV infection, often resulting in insufficient immunological memory and recurrent reinfections (4, 5). Despite its clinical importance, no licensed vaccines or antiviral therapeutics currently exist, underscoring the need for innovative and broadly protective vaccine strategies (6, 7).

HMPV belongs to the Pneumoviridae family and possesses a negative-sense single-stranded RNA genome that encodes multiple structural and non-structural proteins. These include the fusion (F), attachment (G), small hydrophobic (SH), matrix (M2–1 and M2-2), nucleocapsid (N), phosphoprotein (P), and RNA-dependent RNA polymerase (L) proteins, each playing critical roles in viral entry, replication, and immunomodulation. Previous studies have shown that both surface glycoproteins and internal proteins contribute to the induction of humoral and cellular immunity, suggesting that a multi-antigenic approach may be necessary for robust protection (8).

Recent immunoinformatics studies have increasingly explored multi-epitope and reverse-vaccinology strategies for HMPV vaccine development. A recent study used conserved and virulence-associated HMPV proteins to identify B-cell, cytotoxic T-lymphocyte (CTL), and helper T-lymphocyte (HTL) epitopes and incorporated adjuvant and Toll-like receptor (TLR)-related components into a chimeric vaccine construct (9). Other studies have focused on selected HMPV structural or replication-associated proteins. For example, a study investigated Matrix 2-2, Matrix 2-1, Matrix, and phosphoprotein and combined epitope screening with structural modeling, molecular docking, molecular dynamics, immune simulation, and in silico expression analyses (10). Similarly, another study has developed DNA and mRNA vaccine constructs based on six HMPV structural proteins, further demonstrating the growing diversity of computational vaccine-design strategies for this virus (11).

Despite these advances, most reported HMPV immunoinformatics studies have focused on conventional adjuvants, cholera toxin B, or nucleic-acid delivery platforms. The present study explores an alternative computational design strategy by incorporating the B-cell regulatory molecules BAFF and APRIL as immunomodulatory components of multi-epitope constructs. B-cell-activating factor (BAFF) and proliferation-inducing ligand, also known as Tumor necrosis factor ligand superfamily member 13 (APRIL), are members of the TNF superfamily that participate in B-cell survival, differentiation, and antibody responses, providing a biological rationale for investigating their incorporation into vaccine designs (12, 13). The novelty of this work lies in the computational exploration of BAFF- and APRIL-containing multi-epitope HMPV vaccine constructs and their predicted immunological characteristics. In this study, a comprehensive immunoinformatics-driven strategy was employed to design and evaluate multi-epitope vaccine constructs targeting the major structural and non-structural proteins of HMPV, including F, G, N, P, SH, M2-1, M2-2, and L.

2 Materials and methods

2.1 Schematic workflow of HMPV epitope prediction, filtering, and multi-epitope vaccine construction

2.1.1 Retrieval of protein sequences

Amino acid sequences of eight HMPV proteins were selected for the computational vaccine-design workflow: fusion glycoprotein (F), attachment glycoprotein (G), nucleoprotein (N), phosphoprotein (P), small hydrophobic protein (SH), M2-1, M2-2, and RNA-dependent RNA polymerase (L). All HMPV sequences corresponded to the CAN97–83 lineage and were retrieved from GenBank in FASTA format. The corresponding accession numbers and sequence lengths were BBI90015.1 (F, 539 aa), XAT94024.1 (G, 231 aa), AAN52892.1 (N, 394 aa), AAN52865.1 (P, 294 aa), AAQ67698.1 (SH, 179 aa), AAQ67696.1 (M2-1, 187 aa), AAQ67697.1 (M2-2, 71 aa), and AAQ67700.1 (L, 2,005 aa), respectively (Supplementary Table S1). These proteins were selected to provide broad representation of HMPV surface, structural, replication-associated, and accessory proteins as potential sources of T-cell and B-cell epitopes. The amino acid sequences of BAFF and APRIL, which were incorporated as candidate immunomodulatory components of the multi-epitope vaccine constructs, were retrieved from the UniProt database. The BAFF sequence was obtained from UniProt accession Q9Y275 (TN13B_HUMAN) and comprised 285 amino acids, whereas the APRIL sequence was obtained from UniProt accession O75888 (TNF13_HUMAN) and comprised 250 amino acids.

2.1.2 Prediction of physicochemical characteristics

The physicochemical characteristics of the selected HMPV proteins and the final multi-epitope vaccine constructs were evaluated using the ExPASy ProtParam server. The analyzed parameters included molecular weight, theoretical isoelectric point (pI), amino acid composition, numbers of positively and negatively charged residues, instability index, aliphatic index, grand average of hydropathicity (GRAVY), extinction coefficient, and predicted half-life. The ProtParam instability index was interpreted using the commonly applied threshold, whereby values below 40 indicate a theoretically stable protein, whereas values above 40 indicate a potentially unstable protein. The calculated parameters were used to characterize the theoretical physicochemical properties of the proteins and vaccine constructs, including predicted stability, hydropathic properties, and aliphatic composition. These values were interpreted as computational descriptors and not as experimental measurements of protein expression, solubility, folding, thermostability, or biological function.

2.1.3 Computational screening and selection criteria

Candidate epitopes identified through CTL, HTL, B-cell, and IFN-γ prediction analyses were subjected to predefined computational screening and selection procedures (Figure 1). VaxiJen v2.0 was used to predict antigenicity, with a threshold of 0.4. AllergenFP was used to predict allergenic potential, and peptides predicted to be allergenic were excluded. ToxinPred was used for toxicity prediction using the Hybrid (ET+MERCI) model with a threshold of 0.9, and peptides predicted to be toxic were excluded. Potential IFN-γ-inducing epitopes were predicted using the IFNepitope server. The “IFN-gamma versus non-IFN-gamma” prediction model with the motif-and-SVM hybrid algorithm was selected for this analysis. Epitopes predicted to induce IFN-γ were prioritized for subsequent evaluation. The computational screening criteria were applied sequentially to prioritize candidate epitopes according to their predicted antigenicity, allergenicity, toxicity, and IFN-γ-inducing potential. These predictions were used exclusively for candidate prioritization and filtering and were not interpreted as experimental evidence of antigenicity, immunogenicity, allergenicity, toxicity, or cytokine induction. The computational tools, versions, parameters, thresholds, and selection criteria used in the study are summarized in Supplementary Table S2.

2.1.4 CTL epitope prediction and MHC I binding affinity analysis

CTL epitopes were predicted using NetCTL v1.2 and the IEDB analysis resources. Candidate peptides were evaluated based on their predicted MHC class I binding characteristics. In the NetCTL analysis, epitopes meeting the predefined consensus-score criterion used in the present study were prioritized, with a consensus-score threshold of 2. For MHC class I peptide analysis, the full HLA reference set available through the IEDB server was used. Candidate peptides were subsequently screened according to the predefined antigenicity, allergenicity, and toxicity criteria described in Section 2.3. Because the proposed vaccine is intended for human HMPV prevention, human HLA class I alleles were used as the primary MHC restriction elements for CTL epitope selection. Human HLA restriction was distinguished from murine MHC restriction throughout the analysis. The resulting computational predictions were used to prioritize candidate CTL epitopes and do not constitute experimental evidence of MHC binding, T-cell recognition, or cytotoxic activity.

2.1.5 HTL epitope prediction and MHC class II binding analysis

HTL epitopes were evaluated using NetMHCIIpan v4.0 and the IEDB MHC class II prediction resource as complementary computational screening approaches. The peptides ranged from 10 to 15 amino acids in length (10-15-mers). For NetMHCIIpan v4.0, a percentile rank of ≤1% was used to define strong binders, whereas a percentile rank of ≤5% was used to define weak binders. In parallel, the IEDB MHC class II prediction resource was used with its default peptide-length setting and the full HLA reference set available through the IEDB server. The two computational approaches were used as complementary screening procedures to prioritize candidate HTL epitopes. Peptides demonstrating favorable antigenicity and lacking predicted allergenic or toxic properties were retained for construction of the multi-epitope vaccine.

2.1.6 B-cell epitope prediction

Linear B-cell epitopes were predicted using the IEDB B-cell epitope prediction tool. The selected prediction method and default server parameters were used. Predicted B-cell epitopes were subsequently screened according to the predefined computational criteria described in Section 2.3. Peptides satisfying the antigenicity criterion and lacking predicted allergenic or toxic properties were prioritized for incorporation into the multi-epitope vaccine construct. These predictions were interpreted as computational estimates of potential linear B-cell epitope characteristics and were not considered experimental evidence of antibody binding or B-cell immunogenicity.

2.1.7 Prediction of interferon-gamma–inducing epitopes

Potential IFN-γ-inducing epitopes were predicted using the IFNepitope server. The “IFN-gamma versus non-IFN-gamma” prediction model with the motif-and-SVM hybrid algorithm was used. Peptides predicted to be IFN-γ-inducing were subsequently evaluated according to the predefined antigenicity, allergenicity, and toxicity criteria described in Section 2.3. Candidate peptides satisfying these criteria were prioritized for further analysis and vaccine construction. The IFNepitope results were interpreted exclusively as computational predictions of potential IFN-γ-inducing capacity and were not considered experimental evidence of cytokine secretion, T-cell activation, or antiviral immunity.

2.1.8 HLA population coverage analysis

The predicted population coverage of the selected human HLA-restricted T-cell epitopes was evaluated using the IEDB Population Coverage tool. The analysis estimated the proportion of individuals within predefined geographic populations predicted to carry at least one HLA allele associated with the selected epitope set. Population coverage was assessed for Central Africa, East Africa, East Asia, Europe, North Africa, North America, Northeast Asia, and South America. The analysis was used to estimate the breadth of HLA representation among the selected human T-cell epitopes and their potential applicability across genetically diverse populations. The resulting population coverage estimates were interpreted as computational predictions based on the HLA frequencies incorporated into the IEDB database and were not considered direct evidence of population-level vaccine effectiveness or protective immunity.

2.1.8 Multi-epitope vaccine construction and characterization

The multi-epitope vaccine constructs were assembled by linking the selected CTL, HTL, and B-cell epitopes using predefined linker sequences. CTL epitopes were joined using the AYY linker, HTL epitopes were connected using the GPGPG linker, and selected B-cell epitopes were linked using the KK linker. These linkers were incorporated to provide separation between adjacent epitope sequences and to facilitate the intended organization of the multi-epitope construct. BAFF or APRIL was incorporated at the N-terminus of the respective vaccine construct as a candidate immunomodulatory component. Each immunomodulatory component was connected to the epitope region using the EAAAK linker. The resulting construct architecture was subsequently evaluated using the computational analyses described below. The selected linker sequences and BAFF/APRIL components were incorporated as design features of the vaccine constructs; however, their ability to improve epitope processing, structural stability, immunogenicity, or biological activity was not experimentally evaluated in the present study.

2.1.9 Evaluation of the vaccine construct

The final multi-epitope vaccine constructs were evaluated for their predicted antigenic, allergenic, toxicological, and physicochemical characteristics using the computational tools described above. Antigenicity was predicted using VaxiJen v2.0, allergenic potential was assessed using AllergenFP, and toxicity was evaluated using ToxinPred. The physicochemical properties of the constructs were further analyzed using the ExPASy ProtParam server, including molecular weight, theoretical isoelectric point, instability index, and other calculated parameters described in Section 2.2. These analyses were used to characterize and prioritize the vaccine constructs according to predefined computational criteria. The predicted non-allergenic and non-toxic profiles were interpreted as computational outputs and were not considered evidence of clinical safety, tolerability, absence of off-target effects, or biological compatibility.

2.1.10 Secondary structure prediction of the multi-epitope vaccine constructs

The secondary structure of the multi-epitope vaccine construct was predicted using the GOR4 server. GOR4 is based on the Directional Probability Matrix (DPM) algorithm, which predicts protein secondary structural elements by analyzing amino acid sequence information and residue-specific probabilities. The prediction approach incorporates statistical parameters derived from the Chou–Fasman method to identify α-helices, β-sheets, and random coil regions within the protein sequence (14). Additionally, the secondary structure of the multi-epitope vaccine construct was predicted using the PSIPRED server. PSIPRED employs a highly accurate three-state prediction algorithm to classify protein regions into α-helices, β-strands, and coils. The server utilizes position-specific iterative analysis through PSI-BLAST to identify homologous sequences with significant similarity to the vaccine construct, thereby improving the accuracy of secondary structure prediction.

2.1.11 Molecular docking of the vaccine construct with TLR1 and TLR4 receptors

Molecular docking was performed to explore the predicted structural interactions between the multi-epitope vaccine constructs and selected innate immune receptors. TLR1 (UniProt ID: Q15399) and TLR4 (UniProt ID: O00206) were selected as representative receptor targets for this exploratory structural analysis. Three-dimensional structures of the vaccine constructs and TLR1/TLR4 were generated using AlphaFold, and protein–protein docking was subsequently performed using HADDOCK with the default server settings unless otherwise specified. The resulting docking complexes were evaluated based on the predicted interaction interfaces and intermolecular contacts. Because molecular docking provides a computational prediction of possible protein–protein association geometries, the resulting complexes were interpreted as putative structural interactions and not as experimental evidence of receptor binding, binding affinity, receptor activation, downstream signaling, or immunological function. TLR4 was included because previous experimental studies have implicated TLR4 in innate immune responses to HMPV. TLR1 was included as an exploratory Toll-like receptor target based on its established role in innate immune recognition. However, TLR1 generally functions as a heterodimer with TLR2, particularly in the recognition of triacylated lipopeptides. Therefore, docking to isolated TLR1 was considered a simplified structural model and not a representation of the complete physiological TLR1/TLR2 signaling complex. Importantly, TLR1 and TLR4 were not considered cognate receptors for BAFF or APRIL, and the docking analysis was not intended to establish BAFF- or APRIL-mediated signaling through these receptors. BAFF-R, TACI, and BCMA, which represent established receptors within the BAFF/APRIL system, were considered biologically relevant targets for future experimental investigation.

2.1.12 Immune system simulation

The predicted immune-response profile of the designed multi-epitope vaccine was investigated using the C-ImmSim server. The resulting simulations were used to explore theoretical patterns of antibody, B-cell, and T-cell responses and cytokine production and should not be interpreted as experimental evidence of vaccine immunogenicity or protection. The simulation was performed using a random seed of 12,345, a simulation volume of 10, and 100 simulation time steps. The host HLA profile included HLA class I alleles HLA-A01:01, HLA-A02:01, HLA-B07:02, and HLA-B08:01, together with HLA class II alleles HLA-DRB101:01 and HLA-DRB103:01. These alleles were selected as representative HLA class I and class II profiles supported by the simulation framework and were used to examine the predicted immune response under a defined computational HLA background. They should not be considered representative of the full diversity of HLA alleles in the human population. A single vaccine administration was modeled at simulation time step 2 using 1000 antigen units and an adjuvant concentration of 100. These values were used as standardized computational input parameters to examine the predicted kinetics of the immune response rather than as experimentally determined or clinically optimized vaccine doses. The use of a single administration provided a defined computational condition for assessing the simulated primary immune response and subsequent development of memory-cell populations. The simulation was followed for 100-time steps to characterize the temporal behavior of the modeled immune response, including antigen clearance, antibody production, B-cell and T-cell populations, cytokine responses, and predicted memory-cell populations. The simulation duration represents a computational observation period and should not be interpreted as a direct estimate of the duration of protective immunity in humans. The C-ImmSim outputs were interpreted exclusively as computational predictions of potential immune-response patterns. Predicted antibody production, lymphocyte expansion, cytokine secretion, and memory-cell accumulation were not considered experimental evidence of immunogenicity or protective efficacy.

3 Results

3.1 Viral protein sequence characteristics and antigenicity analysis

Amino acid sequences of eight HMPV proteins F, G, N, P, SH, M2-1, M2-2, and L were retrieved from GenBank and used in the subsequent computational analyses. Each sequence was screened for potential immunogenicity and antigenicity using the VaxiJen v2.0 server (http://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html). Amino acid sequences of the eight selected HMPV proteins F, G, N, P, SH, M2-1, M2-2, and L were retrieved from GenBank and used in the subsequent computational analyses. Each sequence was screened for predicted antigenicity using the VaxiJen v2.0 server. The nucleocapsid protein (N; 394 amino acids) demonstrated the highest predicted antigenicity score (0.5286), followed by phosphoprotein P, RNA-dependent RNA polymerase L (0.4770), fusion glycoprotein F (0.4711), attachment glycoprotein G (0.4710), M2-1 (0.4163), small hydrophobic protein SH (0.3781), and M2-2 (0.3630).

3.2 Analysis of physicochemical properties of the target viral proteins

The physicochemical properties of the major HMPV proteins were analyzed using ProtParam (Table 1). The N protein had a molecular weight of 43,537.62 Da, a theoretical pI of 6.68, and a GRAVY value of -0.219, indicating hydrophilicity. The P protein had a molecular weight of 32,747.98 Da, a pI of 4.81, and a GRAVY value of -0.749, indicating high hydrophilicity. The L protein, the largest protein analyzed (230,759.63 Da), had a pI of 8.95 and a GRAVY value of -0.222, consistent with moderate hydrophilicity. The SH protein had a molecular weight of 20,627.71 Da, a pI of 8.62, and a GRAVY value of -0.377, indicating moderate hydrophilicity. The M2–1 protein had a molecular weight of 21,234.09 Da, a pI of 9.14, and a GRAVY value of -0.582, indicating hydrophilicity. In contrast, the smaller M2–2 protein (8,188.64 Da) had a pI of 5.83 and a GRAVY value of 0.092, indicating a slightly hydrophobic character. The F protein had a molecular weight of 58,574.31 Da, a pI of 7.84, and a GRAVY value of 0.028, reflecting a nearly neutral hydropathic profile. The G protein had a molecular weight of 25,445.23 Da, the highest pI (9.90), and the lowest GRAVY value (-0.963), indicating the highest predicted hydrophilicity among the proteins analyzed.

Table 1. Physicochemical properties of the target viral proteins.

3.3 Cytotoxic (CD8⁺) T-cell epitope prediction and MHC class I binding analysis

CTL epitopes were initially predicted from the analyzed HMPV proteins using NetCTL v1.2 and the IEDB analysis resources. Candidate peptides were evaluated according to their predicted MHC class I binding characteristics, and NetCTL candidates with consensus scores below 2 were prioritized according to the selection criterion used in the present study. Eight CTL epitopes (Table 2), representing the eight analyzed viral proteins, were prioritized for incorporation into the multi-epitope vaccine construct. The selected peptides were predicted to have favorable antigenicity and to be non-allergenic and non-toxic according to the applied computational tools. The final CTL epitope panel comprised M2-1 (SNYLLNQL), M2-2 (IIYVKAYL), RNA-dependent RNA polymerase (SIKIYLQL), fusion glycoprotein F (SSFDPVKF), attachment glycoprotein G (LIIDHATL), nucleocapsid N (LSYKHAIL), phosphoprotein P (QEDDIYQL), and small hydrophobic protein SH (ITQYFKRL).

Table 2. Final computationally selected CTL, HTL, and B-Cell epitopes used in the multi-epitope HMPV vaccine construct.

3.4 Helper T lymphocyte epitope prediction and MHC class II binding analysis

HTL epitopes were initially predicted using NetMHCIIpan v3.2 and the IEDB MHC class II prediction resources. Peptides ranging from 10 to 15 amino acids were evaluated. For NetMHCIIpan v3.2, peptides with percentile ranks of ≤1% and ≤5% were classified as strong and weak binders, respectively. The IEDB analysis was performed using the default peptide-length setting and the full HLA reference set available through the IEDB server. Eight HTL epitopes (Table 2), representing the eight analyzed viral proteins, were prioritized for incorporation into the multi-epitope vaccine construct. The selected sequences were M2-1 (KQLQEVEVRQARD), M2-2 (DGIVKSHTNIY), RNA-dependent RNA polymerase (NEEEVGIIAENITPV), fusion glycoprotein F (ATAAAVTAGVA), attachment glycoprotein G (TTENSTSPAAAP), nucleocapsid N (EIGIQYISTALGSER), phosphoprotein P (NIATAGPTAARDGIR), and small hydrophobic protein SH (ITITINYIKVENNLQ).

3.5 B-cell epitopes prediction

Linear B-cell epitopes were initially predicted from the analyzed HMPV protein sequences using the IEDB B-cell epitope prediction resource. The resulting candidate peptides were subsequently screened according to the predefined computational criteria described in the Methods. Predicted antigenicity was assessed using VaxiJen v2.0, while AllergenFP and ToxinPred were used to exclude candidates predicted to be allergenic or toxic, respectively. Eight B-cell epitopes (Table 2), representing the eight analyzed viral proteins, were selected for incorporation into the multi-epitope vaccine construct. The final B-cell epitope panel comprised M2-1 (AGADNDSSYALQDSESINQ), M2-2 (MIWTQKELKEA), RNA-dependent RNA polymerase (NGDNQSIDVS), fusion glycoprotein F (LTCADGPSL), attachment glycoprotein G (STASPSTKPNSQ), nucleocapsid N (LKRYPRMDI), phosphoprotein P (GPTAARDG), and small hydrophobic protein SH (KKIIKDHSG).

3.6 Prediction of interferon-gamma-inducing epitopes

In addition to the B-cell-stimulating epitopes, the analysis identified eight HMPV protein epitopes capable of inducing interferon-gamma (IFN-γ), which were selected for further investigation. Four IFN-γ-inducing epitopes, SIKIYLQL, EIGIQYISTALGSER, MIWTQKELKEA, and LKRYPRMDI (Table 2), representing the eight analyzed viral proteins, were selected for incorporation into the multi-epitope vaccine construct. The final computationally prioritized epitope set is summarized in Table 2, which identifies the selected CTL, HTL, and B-cell epitopes according to their respective HMPV source proteins.

3.7 HLA population coverage of selected T-cell epitopes

The population coverage of the selected human HLA-restricted epitopes was evaluated using the IEDB Population Coverage tool. As shown in Table 3, the predicted population coverage was 100% for all eight geographic populations examined, including Central Africa, East Africa, East Asia, Europe, North Africa, North America, Northeast Asia, and South America. The average number of recognized epitope/HLA combinations ranged from 97.54 in North Africa to 160.61 in Europe, with an overall average of 120.61. The minimum number of epitope/HLA combinations recognized by 90% of the population ranged from 83.20 in North Africa to 141.85 in Europe, with an overall value of 104.14. These results indicate broad population coverage across different geographic populations.

Table 3. Predicted population coverage across different geographic populations.

3.8 Multi-epitope vaccine sequence construction and design

Based on the sequential computational screening of epitopes derived from the F, G, M2-2, M2-1, L, N, P, and SH proteins, a multi-epitope vaccine construct was designed using the selected CTL, HTL, and B-cell epitopes. The final construct comprised 315 amino acid residues and incorporated 24 non-overlapping epitopes, including 8 CTL, 8 HTL, and 8 B-cell epitopes, with one epitope from each of the eight analyzed HMPV proteins for each epitope category. The epitopes were selected according to the predefined computational criteria for MHC binding/prediction, antigenicity, allergenicity, and toxicity described in the Methods. To generate the chimeric construct, different peptide linkers were incorporated between the epitope groups. CTL epitopes were connected using Ala-Ala-Tyr (AAY) linkers, HTL epitopes using Gly-Pro-Gly-Pro-Gly (GPGPG) linkers, and B-cell epitopes using Lys-Lys (KK) linkers. These linkers were selected based on their commonly used properties in multi-epitope vaccine design and were intended to provide separation between adjacent epitopes and facilitate their computationally predicted structural organization. The resulting sequence architecture is shown in Figure 2.

3.9 Physicochemical, antigenic, and allergenic properties of the BAFF-containing vaccine construct

The final multi-epitope vaccine construct exhibited favorable predicted antigenicity and was predicted to be non-allergenic and non-toxic (Table 1). The BAFF-adjuvanted construct comprised 533 amino acids, with a calculated molecular weight of 57.8 kDa and a theoretical pI of 8.88, consistent with its basic character. It contained 62 positively charged residues (Arg + Lys) and 54 negatively charged residues (Asp + Glu). The construct was predicted to be stable, with an instability index of 32.98 and an aliphatic index of 80.28, while its GRAVY value of −0.455 indicated overall hydrophilicity. The predicted extinction coefficient was 47,135 M⁻¹ cm⁻¹. The estimated half-lives were 4.4 h in mammalian reticulocytes (in vitro), >20 h in yeast (in vivo), and >10 h in Escherichia coli (in vivo).

3.10 Secondary structure prediction of the multi-epitope vaccine constructs

The secondary structure prediction generated by the GOR IV server revealed that the multi-epitope vaccine constructs exhibited a predominance of random-coil regions, along with substantial proportions of α-helices and extended β-strands. The vaccine construct containing APRIL as an adjuvant consisted of 50.00% random coils, 32.04% α-helices, and 17.96% extended β-strands Figure 2. In comparison, the second construct with BAFF as an adjuvant exhibited 46.49% random coils, 39.47% α-helices, and 14.04% extended β-strands, as illustrated in Figure 4. The high proportion of random coil regions suggests enhanced structural flexibility, which may facilitate epitope exposure and improve immune recognition. Meanwhile, the presence of α-helices and extended β-strands contributes to the structural stability, proper folding, and conformational organization of the vaccine constructs. The graphical representation showing the positional distribution and probability of the predicted secondary structural elements along the amino acid sequence is presented in Figures 3 and 4.

3.11 Physicochemical, antigenic, and allergenic characterization of the BAFF- and APRIL-containing vaccine construct

The physicochemical properties of the BAFF- and APRIL-adjuvanted multi-epitope vaccine constructs were evaluated using the ProtParam server, while antigenicity, allergenicity, and toxicity were assessed using VaxiJen v2.0, AllergenFP v1.1, and ToxinPred, respectively (Table 4). Both constructs were predicted to be antigenic, non-allergenic, and non-toxic. The BAFF-adjuvanted construct comprised 533 amino acids with a predicted molecular weight of 57.8 kDa, whereas the APRIL-adjuvanted construct comprised 496 amino acids with a molecular weight of 53.63 kDa. Their theoretical pI values were 8.88 and 9.55, respectively, indicating predominantly basic characteristics. Consistent with these values, the BAFF-containing construct had 62 positively charged and 54 negatively charged residues, whereas the APRIL-containing construct had 63 positively charged and 43 negatively charged residues. Both constructs were enriched in alanine, glycine, isoleucine, leucine, and lysine, with low cysteine contents of 1.1% and 0.6% for the BAFF- and APRIL-containing constructs, respectively. Both constructs showed predicted stability, with instability indices below 40 (32.98 for BAFF and 31.76 for APRIL), and relatively high aliphatic indices (80.28 and 81.69, respectively). Their GRAVY values (-0.455 and -0.425) indicated an overall hydrophilic character, which may be compatible with aqueous solubility. The predicted half-lives were identical for both constructs: 4.4 h in mammalian reticulocytes (in vitro), >20 h in yeast (in vivo), and >10 h in Escherichia coli (in vivo). The predicted extinction coefficients were 47,135 and 46,760 M⁻¹ cm⁻¹ for the BAFF-containing construct and 50,895 and 50,770 M⁻¹ cm⁻¹ for the APRIL-containing construct under oxidized and reduced cysteine conditions, respectively. These properties may be compatible with recombinant production; however, experimental studies are required to confirm expression yield, solubility, folding, stability, and functional activity.

Table 4. Comparative physicochemical properties of the BAFF- and APRIL-adjuvanted multi-epitope vaccine constructs.

3.12 Molecular docking analysis of the multi-epitope vaccine construct with TLR4

Molecular docking predicted a putative interaction between the multi-epitope vaccine construct and TLR4, with multiple predicted intermolecular contacts at the receptor–ligand interface (Figure 5, Table 5). The docking pose indicated structural compatibility between the vaccine construct and the leucine-rich repeat (LRR) region of TLR4, with the construct positioned along the lateral surface of the LRR domain. Several hydrogen-bonding, polar, electrostatic, hydrophobic, and van der Waals contacts were identified within the predicted interface. Hydrogen-bonding interactions involved Gln185 and Gln187 of the vaccine construct and Asn622 and Ser623 of TLR4, while additional predicted contacts involved Asp196, Asn575, Glu30, Arg154, Lys184, Pro27, Phe62, and Gly183. Additional peripheral contacts were also identified. Overall, these findings indicate a predicted structural compatibility and putative intermolecular interaction between the vaccine construct and TLR4. However, docking alone cannot establish binding affinity, physiological receptor engagement, receptor activation, or functional signaling, and these predictions require experimental validation.

Table 5. Key residue interactions in the docked complex.

3.13 Molecular docking analysis of the multi-epitope vaccine construct with TLR1

Molecular docking predicted a putative interaction between the HMPV multi-epitope vaccine construct and the modeled TLR1 structure, with multiple predicted non-covalent contacts at the receptor–ligand interface (Figure 6, Table 6). The docking configuration indicated predicted structural compatibility between the vaccine construct and the TLR1 leucine-rich repeat (LRR) region. Hydrogen-bonding, polar, hydrophobic, van der Waals, and electrostatic contacts were identified within the predicted interface. These findings represent computational predictions of structural compatibility and putative intermolecular contacts and do not establish physiological receptor binding, binding affinity, receptor activation, or downstream immune signaling.

Table 6. Key residue interactions in the docked complex.

3.14 In silico immune response simulation

The C-ImmSim simulation was used to predict the theoretical immune-response profile of the designed vaccine constructs following a single modeled immunization. The outputs represent simulated response patterns generated under predefined model assumptions and parameters and therefore should not be interpreted as experimental evidence of immunogenicity, protective efficacy, or long-term immune memory.

3.14.1 Antibody response

Following the modeled vaccine administration, the simulated antigen concentration increased sharply, reaching approximately 7 × 10⁵ antigen counts/mL within the first 2–3 days, followed by a rapid decline to near-undetectable levels by approximately day 5 (Figure 7A). In the simulation, IgM appeared first and peaked around days 10–12 before declining. A subsequent class-switched antibody response was predicted, with IgG1 representing the predominant isotype, followed by IgG1 + IgG2 and IgG2. Simulated IgG levels remained elevated during the modeled observation period, whereas IgE production remained negligible (Figure 7A). These results indicate a predicted humoral-response pattern within the C-ImmSim model rather than experimentally demonstrated antibody production or durable humoral immunity.

3.14.2 Cytokine and interleukin responses

The simulation predicted a transient cytokine response following antigen exposure. IFN-γ showed the highest simulated response, reaching approximately 4 × 10⁵ units around day 4 before declining, while IL-2 also showed a prominent early peak (Figure 7B). Moderate predicted increases were observed for IL-10, IL-12, TNF-α, IL-6, and other cytokines, whereas IL-4, IL-18, and IL-23 remained comparatively low. Overall, the simulated cytokine profile was characterized by relatively prominent IFN-γ and IL-2 responses, consistent with a computationally predicted Th1-associated pattern that may be relevant to antiviral cellular immunity. However, these predictions do not establish a functional Th1 response or antiviral efficacy in vivo.

3.14.3 B-cell population

The simulated total B-cell population increased from nearly zero to approximately 470–480 cells/mm³ by day 6 and remained relatively stable thereafter (Figures 7C, D). The computational model predicted a progressive increase in the memory B-cell population, reaching approximately 280 cells during the modeled observation period. IgG1-producing B cells represented the predominant simulated activated B-cell subset, whereas IgM- and IgG2-producing populations remained comparatively low. Following antigen exposure, active B cells increased during the early simulation phase and subsequently declined as the simulated antigen concentration decreased. Internalized antigen-presenting B cells became the predominant population during the later simulation period. Duplicating and presenting B-cell populations showed relatively small and transient changes, while simulated anergic B cells remained negligible. These findings represent predicted B-cell population dynamics within the computational model and should not be interpreted as evidence of experimentally established memory B-cell formation or long-term humoral immunity.

3.14.4 T-helper (CD4⁺) cell population

The simulated total T-helper cell population increased markedly following the modeled immunization, reaching approximately 5,500–5,600 cells/mm³ by days 8–10 before gradually declining (Figures 7E, F). Non-memory T-helper cells showed a similar pattern, whereas simulated memory T-helper cells progressively increased to approximately 400–450 cells. The active T-helper population increased during the early simulation phase and subsequently declined, while resting T-helper cells became predominant. Duplicating T-helper cells showed a transient increase, whereas anergic T-helper cells remained low throughout the modeled period. These results indicate predicted CD4⁺ T-cell population dynamics within the C-ImmSim framework and do not demonstrate sustained T-cell activation, functional memory, or absence of immune suppression in vivo.

3.14.5 Cytotoxic T cell total population

The simulated total T-cell population remained relatively stable, ranging from approximately 1,095 to 1,127 cells/mm³ (Figures 7G, H). Memory and non-memory T-cell populations showed relatively limited variation following the modeled single immunization. The simulated active T-cell population increased progressively, whereas resting T cells gradually declined. Duplicating T cells remained minimal, and anergic T cells were essentially absent during the simulation period. These findings describe predicted T-cell dynamics within the computational model and should not be interpreted as experimental evidence of sustained T-cell activation, functional immune memory, or absence of immune dysfunction in vivo.

4 Discussion

This study presents an immunoinformatics framework for the rational design of a multi-epitope vaccine against HMPV. Although HMPV is a major cause of acute respiratory tract infections, particularly in infants, older adults, and immunocompromised individuals, no licensed vaccine is currently available. Viral genetic diversity, immune evasion, and the lack of well-defined correlates of long-term protective immunity remain challenges for vaccine development (15–18). By integrating antigenicity screening, epitope prediction, vaccine construction, physicochemical characterization, and molecular docking, this study identified candidate epitopes from multiple structural and non-structural HMPV proteins and incorporated them into BAFF- and APRIL-adjuvanted multi-epitope vaccine constructs with favorable predicted immunological and physicochemical properties.

Recent studies have demonstrated the utility of immunoinformatics and reverse-vaccinology approaches for HMPV vaccine development. Multi-epitope vaccine candidates targeting HMPV alone or in combination with other respiratory viruses have been designed using computational epitope prediction, antigenicity and safety screening, structural modeling, molecular docking, and immune simulation (9, 19, 20). In particular, recent studies have explored HMPV-based multi-epitope constructs and broader respiratory-virus vaccine designs, supporting the feasibility of computational approaches for identifying conserved epitopes and constructing candidate vaccines (9, 21). The present study extends these approaches by integrating epitopes derived from eight HMPV proteins (F, G, M2-1, M2-2, L, N, P, and SH) and incorporating BAFF or APRIL as immunomodulatory components. However, consistent with the limitations of computational vaccine studies, the predicted immunological and structural properties require experimental validation before conclusions regarding immunogenicity or protective efficacy can be established. A distinguishing feature of the present study is the incorporation of BAFF and APRIL as candidate immunomodulatory components. BAFF and APRIL were selected as candidate immunomodulatory components because of their established roles in B-cell survival, differentiation, and antibody responses. These properties provide a rationale for exploring their incorporation into a multi-epitope vaccine construct; however, their effects in the present vaccine design remain computationally predicted and require experimental validation. BAFF and APRIL were incorporated into the vaccine constructs as potential immunomodulatory components based on their established roles in B-cell survival, differentiation, and antibody production. However, their effects depend on receptor engagement, concentration, exposure duration, and immunological context (22–24). Experimental vaccine studies have also explored BAFF and APRIL as molecular adjuvants, with evidence of enhanced germinal-center and antigen-specific antibody responses in animal models (25). However, excessive or dysregulated BAFF/APRIL signaling can alter B-cell homeostasis and impair immune tolerance, with increased BAFF activity associated with the persistence of autoreactive B cells and autoimmune disease (26–28). Therefore, the incorporation of BAFF or APRIL should not be assumed to confer improved safety or reduced reactogenicity compared with conventional adjuvants (29, 30). Excessive or prolonged signaling may disrupt B-cell homeostasis and immune regulation. Accordingly, the present computational findings do not establish that BAFF- or APRIL-containing constructs are safer or superior to established adjuvants (22–24).In the present study, the predicted non-allergenic and non-toxic profiles provide only preliminary computational information and do not establish clinical safety, tolerability, or absence of off-target immunological effects. Therefore, BAFF and APRIL should not be considered equivalent to conventional vaccine adjuvants. Experimental studies will be required to determine the appropriate dose, biological activity, inflammatory effects, effects on B-cell homeostasis, and potential autoimmune or other off-target consequences of the proposed constructs.

The molecular docking analyses predicted putative interactions between the designed vaccine construct and the innate immune receptors TLR1 and TLR4. The predicted complexes contained multiple hydrogen-bonding, electrostatic, hydrophobic, and van der Waals contacts, indicating structural compatibility within the modeled receptor-ligand interfaces. However, these computational observations should not be interpreted as evidence of receptor activation or functional signaling. Although TLRs are important mediators of innate immune responses, the biological consequences of the predicted interactions remain unknown and require experimental investigation (31–37). The docking analyses predicted putative receptor–ligand interactions involving hydrogen bonds, electrostatic contacts, hydrophobic interactions, and salt bridges. These findings indicate predicted structural compatibility between the vaccine construct and the leucine-rich repeat (LRR) domains of TLR1 and TLR4 (34, 35, 38–40). In the TLR4 complex, residues including Gln185, Gln187, Glu30, Pro27, and Asp50 contributed to the predicted interaction interface, whereas TLR1 docking involved residues such as Gln178, Pro180, Pro182, Lys248, Glu219, and Asn166. However, molecular docking alone cannot establish receptor binding affinity, physiological receptor engagement, receptor activation, or downstream immune signaling. Therefore, these findings should be interpreted as computational predictions of structural compatibility and putative intermolecular contacts rather than evidence of functional receptor engagement or innate immune activation. Experimental binding and functional studies are required to determine whether these predicted interactions occur under physiological conditions. TLR4 was selected based on its reported involvement in HMPV-related innate immunity, while TLR1 was examined as an additional structural target. Because TLR1 functions primarily as a TLR1/TLR2 heterodimer, docking to isolated TLR1 represents a simplified model and does not demonstrate receptor activation or a specific adjuvant mechanism (41, 42). Thus, the docking results should be interpreted as predictions of structural compatibility and putative intermolecular contacts requiring experimental validation. Molecular dynamics simulations and binding free-energy analyses could further assess the stability and energetic properties of the predicted complexes.

The C-ImmSim analysis predicted a coordinated pattern of humoral and cellular immune responses following the modeled single immunization, including simulated antibody production, B- and T-cell population changes, cytokine responses, and accumulation of memory-cell populations. These findings represent model-derived predictions and should not be interpreted as experimental evidence of vaccine immunogenicity or protective efficacy. IFN-γ and IL-2 showed prominent predicted responses during the early simulation phase, consistent with a Th1-oriented cellular response that may be relevant to antiviral immunity against HMPV. IFN-γ is an important mediator of antiviral cellular immunity, while IL-2 promotes T-cell proliferation and expansion (43, 44). Because C-ImmSim outputs depend on predefined model assumptions and parameters, the predicted immune-cell and cytokine kinetics cannot be directly equated with physiological responses following vaccination in humans.

The physicochemical analysis provided preliminary computational characterization of the BAFF- and APRIL-containing vaccine constructs. The predicted instability index, aliphatic index, GRAVY values, molecular weight, and theoretical half-life suggested characteristics potentially compatible with recombinant production. However, these predictions do not establish successful expression, solubility, correct folding, structural integrity, stability, or biological functionality. Experimental recombinant production and subsequent biochemical and immunological characterization will therefore be required to validate these computational findings.

Overall, this study provides a comprehensive computational framework for the rational design and prioritization of next-generation HMPV vaccine candidates. The incorporation of the endogenous immunomodulators BAFF and APRIL resulted in vaccine constructs with favorable predicted antigenicity, stability, safety, and physicochemical properties. Docking analyses further predicted structural compatibility with selected innate immune receptors. Although these findings are computational and require experimental validation, they support further investigation of BAFF- and APRIL-adjuvanted multi-epitope vaccines as candidate approaches for HMPV prevention.

5 Conclusion

This study presents an in silico multi-epitope vaccine design targeting multiple HMPV proteins and incorporating BAFF or APRIL as candidate immunomodulatory components. The selected constructs exhibited favorable predicted antigenic, physicochemical, structural, and immunological properties. Molecular docking predicted putative structural interactions with TLR1 and TLR4, while C-ImmSim predicted humoral and cellular immune-response patterns. Population coverage analysis further indicated broad predicted HLA representation of the selected T-cell epitopes. However, all reported findings are computational predictions and do not establish receptor binding or activation, immunogenicity, safety, antiviral efficacy, or protective immunity. The BAFF- and APRIL-containing constructs should therefore be regarded as preliminary computational vaccine candidates. Experimental studies are required to validate recombinant expression, structural properties, receptor interactions, antigen-specific immune responses, safety, and protective efficacy before their suitability for preclinical development can be assessed.

6 Limitations

Despite the encouraging computational findings, several limitations should be acknowledged. First, this study was based entirely on in silico predictions; therefore, the predicted antigenicity, immunogenicity, allergenicity, toxicity, physicochemical properties, and receptor interactions require experimental validation. Second, molecular docking provides a static representation of putative receptor–ligand interactions and cannot adequately capture conformational flexibility or binding stability under physiological conditions. Moreover, the docking results were interpreted qualitatively based on predicted interaction interfaces and residue-level contacts and should not be considered evidence of functional receptor binding or activation. Molecular dynamics simulations and binding free-energy calculations could provide further insight into the stability and energetic properties of the predicted complexes. Finally, recombinant protein expression, structural and biochemical characterization, in vitro immunological and cytokine assays, in vivo immunogenicity, safety, and protection studies are required to determine whether the proposed vaccine constructs are biologically functional and protective.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The graphical abstract was conceptualized by Dr. Maaweya Awadalla. Generative artificial intelligence (ChatGPT, OpenAI) was used to support visualization during preparation of the graphical abstract. All content was reviewed, verified, and approved by the authors.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fviro.2026.1949694/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Amelia K Pinto, University of Kentucky, United States

Reviewed by: Leonardo Pereira De Araújo, Federal University of Alfenas, Brazil

Maryam Ehsasatvatan, University of Tabriz, Iran