How Peptide Vaccines Are Designed: Epitope to Injection

Peptide Vaccines

30+ Peptide Vaccines in Development

More than 30 peptide-based vaccines are currently in clinical trials, from personalized cancer vaccines to infectious disease targets, all built on epitope prediction and MHC binding.

Ma et al., Scandinavian Journal of Immunology, 2020

Ma et al., Scandinavian Journal of Immunology, 2020

View as image

A peptide vaccine is a synthetic fragment of a protein, typically 8 to 30 amino acids long, designed to train the immune system to recognize a specific target. Unlike whole-pathogen or mRNA vaccines, peptide vaccines deliver only the minimal molecular information needed to trigger an immune response. Ma et al. (2020) reviewed the evolution from universal shared-antigen approaches to personalized neoantigen-based peptide vaccines, documenting over 30 candidates in clinical trials across cancer and infectious disease.^[1] For an overview of delivery platforms, see self-assembling peptide nanoparticles: the new vaccine delivery platform.

Step 1: Target Identification

Every peptide vaccine begins with identifying the right target protein. The target must meet two criteria: it must be present on the cell you want the immune system to attack, and it must be sufficiently different from normal proteins that the immune system can distinguish it.

Cancer Targets

For cancer vaccines, targets fall into two categories:

• Shared tumor-associated antigens (TAAs): proteins overexpressed or abnormally expressed in many cancers. Examples include HER2, gp100, NY-ESO-1, MAGE-A3, and hTERT. These are "off the shelf" targets that apply to many patients. Lilleby et al. (2017) tested an hTERT peptide vaccine (UV1) in metastatic prostate cancer, inducing immune responses in 18 of 22 patients.^[7] For specific clinical histories with TAA-based vaccines, see gp100 peptide vaccine for melanoma and HER2 peptide vaccines for breast cancer.

• Neoantigens: unique mutations in an individual patient's tumor that create novel peptide sequences. These are highly specific and less likely to trigger immune tolerance, but require patient-specific tumor sequencing. Ma et al. (2020) described the shift from universal to personalized peptide vaccines as one of the most significant developments in cancer immunotherapy.^[1]

Infectious Disease Targets

For pathogen vaccines, targets are surface proteins or conserved structural proteins of viruses, bacteria, or parasites. The target must be accessible to the immune system and sufficiently conserved across pathogen strains that the vaccine provides broad protection.

Step 2: Epitope Prediction

Once a target protein is identified, the next step is determining which short fragment (epitope) of that protein will bind MHC molecules and activate T cells. This is where computational biology becomes essential.

MHC Class I Epitopes (CD8+ T Cells)

MHC class I molecules present peptides of 8-11 amino acids to CD8+ cytotoxic T cells. Predicting which peptides will bind a given MHC class I allele is the most mature area of computational immunology. Nilsson et al. (2025) published NetMHCpan-4.2, the latest iteration of the leading MHC-I binding prediction tool. The updated model uses transfer learning (training first on binding data, then refining with actual epitope data) and structural features of the peptide-MHC complex, achieving improved prediction of which peptides will be truly immunogenic, not just MHC-bound.^[5]

MHC Class II Epitopes (CD4+ T Cells)

MHC class II molecules present longer peptides (13-25 amino acids) to CD4+ helper T cells. Prediction is harder because the binding groove is open at both ends, allowing more variation in peptide length and binding register. Buckley et al. (2022) evaluated existing computational models for CD8+ T cell epitope prediction and found that while MHC binding prediction has improved substantially, predicting which MHC-bound peptides actually trigger a T cell response remains a major challenge.^[8]

The Binding-Immunogenicity Gap

A critical limitation: MHC binding does not equal immunogenicity. A peptide can bind MHC with high affinity but still fail to activate T cells. Schmidt et al. (2017) demonstrated this directly, finding "strong divergence between prediction and observation of T-cell-recognized tumor antigen T-cell epitopes." Computational predictions overestimated the number of immunogenic epitopes by a large margin.^[3]

Wells et al. (2020) addressed this gap by assembling a global consortium (TESLA) that tested predictions from 28 teams against experimental data on tumor neoantigen immunogenicity. The consortium identified key parameters that improve neoantigen prediction: peptide-MHC stability (not just binding affinity), expression level of the source gene, and the degree of difference between the mutant and wild-type peptide. Teams that incorporated these parameters outperformed those relying on binding prediction alone.^[2] For a deeper look at the computational tools used in these predictions, see T-cell epitope prediction: using computers to design vaccines. For the broader role of AI in peptide research, see how AI is revolutionizing peptide drug discovery.

Step 3: Peptide Synthesis and Formulation

Once epitopes are selected, they are synthesized chemically using solid-phase peptide synthesis (SPPS). This process is fast (days, not weeks), scalable, and avoids the biological contamination risks of cell-based manufacturing.

The synthetic peptide alone is a poor immunogen. Short peptides are rapidly degraded by proteases in the body and do not trigger the innate immune alarm signals that help activate adaptive immunity. This means every peptide vaccine requires additional components:

• Adjuvants: substances that amplify the immune response. Common choices include Montanide ISA-51 (a water-in-oil emulsion), poly-ICLC (a TLR3 agonist), and GM-CSF (a cytokine that recruits dendritic cells). The choice of adjuvant can make or break a peptide vaccine's efficacy. For a dedicated analysis, see peptide vaccine adjuvants: why the peptide alone is not enough.

• Carrier proteins or nanoparticles: linking the peptide to a larger protein or embedding it in a nanoparticle can improve uptake by antigen-presenting cells and slow degradation. Self-assembling peptide nanoparticles represent one of the newer approaches.

• Multiple epitopes: most modern peptide vaccines include several epitopes targeting different parts of the antigen and binding different MHC alleles. This increases population coverage (different people have different HLA types) and reduces the chance of immune escape through mutation.

Holmberg-Thyden et al. (2022) tested a multi-epitope peptide vaccine targeting four shared tumor antigens (NY-ESO-1, MAGE-A3, PRAME, WT-1) combined with the hypomethylating agent azacitidine in patients with high-risk myelodysplastic syndrome. The combination approach used multiple epitopes to cover diverse HLA backgrounds in the patient population.^[9]

Step 4: Preclinical Testing

Before human trials, peptide vaccines undergo testing in animal models to assess:

• Immunogenicity: does the vaccine generate detectable T cell responses?
• Tumor protection: in cancer models, does vaccination slow or prevent tumor growth?
• Safety: are there off-target immune responses or autoimmune effects?

Animal models have significant limitations for peptide vaccines. Mice have different MHC molecules (H-2 instead of HLA), so a peptide that binds human HLA-A*02:01 may not bind the mouse equivalent. Transgenic mice expressing human HLA alleles partially address this but do not fully recapitulate the human immune repertoire.

Step 5: Clinical Development

Peptide vaccine clinical trials follow the standard phase I/II/III pathway, with some unique considerations.

Tang et al. (2022) reviewed all registered peptide vaccine clinical trials in gynecologic oncology and found that while many candidates showed immune responses in Phase I/II, translating those responses into objective clinical benefit has been challenging. No peptide vaccine has been licensed for cancer treatment.^[6]

Personalized Approaches

The newest generation of peptide vaccines is fully personalized. The workflow:

1. Sequence the patient's tumor and germline DNA
2. Identify somatic mutations
3. Predict neoantigen epitopes based on the patient's HLA type
4. Synthesize a custom multi-peptide vaccine
5. Administer with adjuvant

Heitmann et al. (2024) demonstrated a warehouse-based approach that streamlines this process. Instead of synthesizing each patient's vaccine from scratch, they pre-manufactured a library of peptides covering common tumor antigens across multiple HLA types. The patient's HLA type determines which pre-made peptides are combined into their personal vaccine. In a Phase II trial for chronic lymphocytic leukemia, the approach proved feasible and generated measurable immune responses.^[4]

Huang et al. (2025) reviewed advances in personalized peptide vaccine development for colorectal cancer, noting that combining peptide vaccines with immune checkpoint inhibitors may overcome the immunosuppressive tumor microenvironment that has limited peptide vaccine efficacy as monotherapy.^[10]

Why Peptide Vaccines Have Not Yet Succeeded

Despite decades of research, no standalone peptide vaccine has achieved regulatory approval for cancer treatment. Several factors explain this:

The immunosuppressive tumor microenvironment: tumors actively suppress local immune responses through regulatory T cells, myeloid-derived suppressor cells, and inhibitory checkpoint molecules. A vaccine may generate T cells in the periphery that become inactivated upon entering the tumor.

Immune tolerance to self-antigens: TAA-based vaccines target proteins that are also present (at lower levels) on normal cells. The immune system has mechanisms to prevent attacking self-proteins, and these tolerance mechanisms limit vaccine-induced T cell responses against TAAs.

HLA diversity: the human population has thousands of HLA alleles. A peptide that binds HLA-A02:01 (found in approximately 40-50% of Caucasians) will not bind HLA-A24:02 (common in East Asian populations). Multi-epitope and personalized approaches address this but add complexity.

Endpoint challenges: cancer vaccine trials must show clinical benefit (survival, tumor shrinkage), not just immune responses. Many trials have shown T cell responses without corresponding clinical outcomes, suggesting that immune response magnitude or quality was insufficient. Farrell (2021) developed the epitopepredict tool to improve MHC binding prediction pipelines, noting that better epitope selection is necessary but not sufficient for clinical success.^[11]

Limitations

This article focuses on the design pipeline rather than clinical outcomes, which vary widely by target, adjuvant, and patient population. The field is evolving rapidly: AI-driven epitope prediction, combination with checkpoint inhibitors, and nanoparticle delivery systems are all changing the landscape in ways that published reviews may not yet fully capture.

The computational tools described (NetMHCpan, TESLA parameters) improve prediction accuracy but do not eliminate the need for experimental validation. Every predicted epitope must still be tested in binding assays and T cell activation assays before clinical use.

Personalized neoantigen vaccines are technically feasible but logistically complex and expensive. Manufacturing a custom vaccine for each patient within a clinically relevant timeframe (weeks, not months) remains a practical challenge. The warehouse-based approach described by Heitmann et al. (2024) offers one solution but limits epitope selection to pre-manufactured candidates.^[4]

The Bottom Line

Peptide vaccine design follows a stepwise pipeline: identify the target protein, predict MHC-binding epitopes computationally, synthesize the peptides, formulate with adjuvant, test in animals, and advance through clinical trials. The critical bottleneck is the gap between MHC binding prediction and actual T cell immunogenicity. Global consortia (Wells et al., 2020) and improved tools (NetMHCpan-4.2) are narrowing this gap. Personalized and warehouse-based approaches represent the current frontier, though no peptide vaccine has yet achieved regulatory approval for cancer.

Frequently Asked Questions

How are peptide vaccines designed?

Peptide vaccines are designed through a multi-step pipeline: target protein identification, computational epitope prediction using tools like NetMHCpan, chemical synthesis via solid-phase peptide synthesis, formulation with adjuvants to boost immunogenicity, preclinical testing in animal models, and clinical trials. The key challenge is predicting which short peptide fragments will both bind MHC molecules and activate T cells.

What is an epitope in vaccine design?

An epitope is the specific short fragment of a protein (typically 8-11 amino acids for MHC class I, 13-25 for MHC class II) that is presented to T cells by MHC molecules on the cell surface. Selecting the right epitope is the most critical step in peptide vaccine design because it determines whether the vaccine will generate an effective immune response against the target.

Why do peptide vaccines need adjuvants?

Short synthetic peptides are rapidly degraded by proteases and do not trigger innate immune alarm signals on their own. Adjuvants amplify the immune response by activating dendritic cells, prolonging antigen presentation, and creating a local inflammatory environment that enhances T cell activation. Common adjuvants include Montanide ISA-51, poly-ICLC, and GM-CSF.

Are personalized peptide vaccines available?

Personalized peptide vaccines based on individual tumor mutations are in clinical trials but not yet commercially available. The process involves tumor sequencing, neoantigen prediction based on the patient's HLA type, and custom peptide synthesis. Heitmann et al. (2024) demonstrated a warehouse-based approach that pre-manufactures peptide libraries for faster personalization in a Phase II trial.

Why have peptide cancer vaccines not been approved?

No standalone peptide cancer vaccine has received regulatory approval due to several challenges: the immunosuppressive tumor microenvironment inactivates vaccine-induced T cells, immune tolerance limits responses against self-antigens, HLA diversity restricts which patients respond to any given epitope, and clinical trials have often shown immune responses without corresponding tumor shrinkage or survival benefit.

What is NetMHCpan used for in vaccine design?

NetMHCpan is a computational tool that predicts which peptide fragments will bind to MHC class I molecules. Nilsson et al. (2025) published version 4.2, which uses transfer learning and structural features to predict not just MHC binding but which bound peptides are likely to be true CD8+ T cell epitopes. It is used in most modern peptide vaccine design pipelines to screen candidate epitopes before experimental testing.