Computational Design Yields Two mRNA Vaccine Candidates Against HMPV Glycoprotein

Human metapneumovirus, a leading cause of acute respiratory infections in children, older adults, and immunocompromised people, still has no licensed vaccine or specific antiviral treatment. A new computational study has taken a step toward filling that gap by designing two multi-epitope mRNA vaccine candidates against the virus’s glycoprotein, using a reverse vaccinology pipeline built entirely on immunoinformatic predictions. The work, published in Discover Immunity, reports favorable predicted antigenicity, safety, and structural properties for both constructs, while emphasizing that every result remains computational and awaits experimental confirmation.

The research team, led by Elijah Kolawole Oladipo with collaborators across multiple Nigerian and international institutions, retrieved 107 human metapneumovirus glycoprotein sequences from the National Center for Biotechnology Information protein database to ensure global representation. These sequences, drawn from all available continents, were screened in two stages: first for antigenicity using VaxiJen 2.0, and then for immunogenicity using VaxiJen 3.0. Only sequences with an antigenicity score greater than 0.6 and an immunogenicity score between 66 and 100 percent advanced to epitope prediction.

From the screened sequences, the researchers predicted three classes of epitopes. Cytotoxic T-lymphocyte epitopes were identified with NetCTL v1.2 and NetMHCpan 4.0, then filtered for antigenicity, toxicity, and allergenicity, yielding five final candidates. Helper T-lymphocyte epitopes were predicted through a consensus of the IEDB and NetMHCIIpan 4.0 servers, with the resulting five epitopes additionally evaluated for their capacity to induce interferon-gamma, interleukin-4, interleukin-6, and interleukin-10. Linear B-cell epitopes were predicted using ABCPred and Bepipred, producing six epitopes that passed the same safety and antigenicity filters. To reduce the risk of autoimmune cross-reactivity, all selected epitopes were subjected to BLASTp analysis against the non-redundant human proteome, with stringency thresholds including an E-value cutoff of 0.001, sequence identity below 30 percent, and query coverage under 50 percent; only epitopes showing no significant homology to human proteins were retained.

The final constructs each incorporated fifteen epitopes: five B-cell, five CTL, and five HTL. The design followed established mRNA vaccine architecture, beginning with a 5′ cap and Kozak sequence to promote translation initiation, followed by a tissue plasminogen activator signal sequence for secretion and an MHC I-targeting domain to route CTL epitopes to the endoplasmic reticulum. Human beta defensin 1 served as an adjuvant at the N-terminus, connected to helper T-lymphocyte epitopes via GPGPG linkers chosen for their flexibility and ability to promote proper epitope presentation. AAY linkers separated CTL epitopes to facilitate proteasomal cleavage and improve MHC class I antigen processing, while EAAAK linkers maintained structural separation between B-cell epitope domains. A 3′ untranslated region and poly-A tail were appended to bolster mRNA stability and translational efficiency.

Safety and physicochemical profiling returned encouraging predictions. The antigenicity scores of constructs 1 and 2 were 0.4984 and 0.4980 respectively, which the authors interpret as an ability to produce antibodies. Both constructs were predicted to be non-allergenic and non-toxic. Molecular weights came out at 120160.44 Da and 120306.28 Da, with theoretical pI values of 8.62 and 8.66 indicating a basic nature. In vitro half-life predictions suggested approximately 30 hours in mammalian reticulocytes. Instability indices of 38.589 and 39.22 classified both as stable proteins, while aliphatic indices of 76.97 and 74.82 reflected thermostability. Negative GRAVY scores of -0.418 and -0.466 confirmed hydrophilic character, and solubility predictions returned scores of 0.345 and 0.296, indicating moderate aqueous solubility.

Structural modeling employed AlphaFold 3, SWISS-MODEL, and Phyre2 to predict tertiary structures, followed by refinement with GalaxyRefine. Secondary structure analysis using SOPMA showed construct 1 with 43.66 percent alpha-helix, 10.20 percent extended strands, 0.00 percent beta-turn, and 46.14 percent random coils, while construct 2 exhibited 41.45 percent alpha-helix, 10.20 percent extended strands, 0.00 percent beta-turn, and 48.35 percent random coils. Validation through Ramachandran plot analysis revealed that construct 1 had 95.9 percent of residues in the favored region and construct 2 had 95.0 percent, with only 0.6 and 0.7 percent respectively in disallowed regions. ERRAT quality factors of 97.307 for construct 1 and 96.256 for construct 2 supported the structural reliability of both models. Conformational B-cell epitopes were also mapped onto the three-dimensional structures using ElliPro, identifying six discontinuous epitopes per construct with scores ranging from 0.841 to 0.958.

To assess innate immune recognition, the team docked both constructs against Toll-like receptors 2 and 4, pattern recognition receptors that trigger type I interferon and pro-inflammatory cytokine production. Using the HDock server with crystal structures TLR-2 (PDB ID: 1O77) and TLR-4 (PDB ID: 2Z63), construct 1 produced docking scores of -278.71 with TLR2 and -349.32 with TLR4, while construct 2 scored -222.82 and -309.86 respectively. The authors interpret these strong binding affinities as suggesting efficient immune receptor engagement. Molecular dynamics simulations using the iMOD server, which performs normal mode analysis in internal coordinates, indicated stable receptor-vaccine complexes throughout the simulation period, supported by hydrogen bonds, salt bridges, and key interface residues identified through PDBsum interaction analysis.

Immune response simulation with the C-ImmSim tool modeled the interaction among immune cells, signaling molecules, and the vaccine constructs. The simulation predicted a typical primary immune response followed by an enhanced secondary response upon re-exposure, suggesting the establishment of immunological memory. Significant activation of helper T cells and generation of long-lived memory B cells and T cells were also observed in the model. Separately, codon optimization for expression in Escherichia coli K12 using the Java Codon Adaptation Tool achieved ideal codon adaptation index values of 1.0 with balanced GC content of 51.5 percent and 51.7 percent for constructs 1 and 2 respectively, and in silico cloning into the pET-28a(+) vector incorporated EcoRI and BamHI restriction sites at the termini.

The choice of the glycoprotein rather than the more commonly targeted fusion protein reflects its distinct immunomodulatory role. The HMPV G protein functions as a virulence factor that suppresses innate immune signaling by interfering with RIG-I-dependent gene transcription and MAVS activation in airway epithelial cells, and disrupts TLR4-mediated signaling in dendritic cells, impairing CD4-positive T-cell activation. Targeting this protein may help counteract viral immune evasion mechanisms. The authors note that although the fusion protein remains the principal target of neutralizing antibodies in most HMPV vaccine studies, the glycoprotein’s pathogenic functions make it an attractive complementary target.

The study’s authors are explicit about its limitations: the entire investigation is computational, and the antigenicity, stability, and immunogenicity of the constructs remain to be validated experimentally. They call for further in vitro and in vivo investigations to confirm safety, immunogenicity, and protective efficacy before any preclinical or clinical development can proceed. The work provides a detailed computational foundation and a concrete set of candidate sequences, but the distance between predicted and demonstrated protection against human metapneumovirus remains substantial, and only laboratory and animal studies can determine whether these designs translate into a viable vaccine.

Subject of Research: Immunoinformatic design of multi-epitope mRNA vaccine candidates against the human metapneumovirus glycoprotein

Article Title: Harnessing reverse vaccinology for the design and validation of mRNA vaccine targeting the glycoprotein of human metapneumovirus (HMPV)

Article References: Oladipo, E. K., Ogunniran, J. A., Akinpelu, O. S., Akpan, B. I., Hamzah, H., Ogunmodede, S. P., Oluwasanya, G. J., Adaramola, E. O., Popoola, P. O., Adedokun, P. T., Gbadero, T. S., Jimah, E. M., Awoyelu, E. H., Iwalokun, B. A., Senbadejo, T. Y., Ariyo, O. E., Akinde, S. B., Olaniyan, O. P., Onyeaka, H., & Olatunde, S. K. (2026). Harnessing reverse vaccinology for the design and validation of mRNA vaccine targeting the glycoprotein of human metapneumovirus (HMPV). Discover Immunity, 3(1), Article 15. https://doi.org/10.1007/s44368-026-00027-w

Image Credits: AI Generated

DOI: 10.1007/s44368-026-00027-w

Keywords: human metapneumovirus, HMPV, mRNA vaccine, reverse vaccinology, immunoinformatics, multi-epitope vaccine, glycoprotein, molecular docking, Toll-like receptors, in silico, epitope prediction, vaccine design