Abstract
Background
Gliomas, including glioblastoma (GBM) and low-grade gliomas (LGG), are resistant to immunotherapy, potentially due to an immunosuppressive microenvironment. Neoantigens - tumour-specific, mutated peptides presented on MHC-I - are promising vaccine targets, yet prediction tools have not yet been validated in glioma. We performed the first large-scale benchmarking of neoantigen prediction tools.
Methods
We analysed 1,150 gliomas (634 GBM, 516 LGG), including 211 CPTAC cases with matched WES, RNA-seq, and MHC-I immunopeptidomics. Expressed mutations were translated into 8–11-mer peptides and assigned HLA alleles via OptiType. Peptides were scored and ranked by 30 prediction tools across binding affinity (BA), eluted ligand (EL), and immunogenicity (IM) classes. Immune infiltration was inferred using TIMER2, and TCR recognition predicted via TITAN. To identify recurrent epitopes, we focused on peptides predicted in ≥15% of patients.
Results
Of all mutated peptide candidates assessed, >12,000 were predicted to have MHC-I binding potential. Tool concordance was low: only 12% of top-20 peptides overlapped across >4 tools, with similar discordance at broader thresholds (top-50, top-100). Many top-ranked predictions - especially from IM-based tools - were not MS-validated, highlighting limitations in current models. Despite this, we identified a conserved set of 19 epitopes. Of these, 11 were confirmed by MS, enriched in CD8-infiltrated tumours, and associated with improved survival (multivariate HR = 0.48, p = 0.003), particularly in IDH-wildtype, MGMT-methylated GBM. These epitopes were clonal, showed high TCR recognition potential, and passed manufacturability criteria. Several had previously demonstrated immunogenicity in NeoVax trials. Together, this validated epitope set achieved population coverage in ∼68% of patients, supporting potential utility in glioma vaccine design.
Conclusions
Most tools miss tumour-presented peptides in glioma, risking ineffective vaccine design. Our benchmarking identifies a recurrent, immunogenic, and manufacturable epitope set with broad coverage. Glioma-specific, MS-informed approaches are needed to improve neoantigen selection and accelerate vaccine development.
Gliomas, including glioblastoma (GBM) and low-grade gliomas (LGG), are resistant to immunotherapy, potentially due to an immunosuppressive microenvironment. Neoantigens - tumour-specific, mutated peptides presented on MHC-I - are promising vaccine targets, yet prediction tools have not yet been validated in glioma. We performed the first large-scale benchmarking of neoantigen prediction tools.
Methods
We analysed 1,150 gliomas (634 GBM, 516 LGG), including 211 CPTAC cases with matched WES, RNA-seq, and MHC-I immunopeptidomics. Expressed mutations were translated into 8–11-mer peptides and assigned HLA alleles via OptiType. Peptides were scored and ranked by 30 prediction tools across binding affinity (BA), eluted ligand (EL), and immunogenicity (IM) classes. Immune infiltration was inferred using TIMER2, and TCR recognition predicted via TITAN. To identify recurrent epitopes, we focused on peptides predicted in ≥15% of patients.
Results
Of all mutated peptide candidates assessed, >12,000 were predicted to have MHC-I binding potential. Tool concordance was low: only 12% of top-20 peptides overlapped across >4 tools, with similar discordance at broader thresholds (top-50, top-100). Many top-ranked predictions - especially from IM-based tools - were not MS-validated, highlighting limitations in current models. Despite this, we identified a conserved set of 19 epitopes. Of these, 11 were confirmed by MS, enriched in CD8-infiltrated tumours, and associated with improved survival (multivariate HR = 0.48, p = 0.003), particularly in IDH-wildtype, MGMT-methylated GBM. These epitopes were clonal, showed high TCR recognition potential, and passed manufacturability criteria. Several had previously demonstrated immunogenicity in NeoVax trials. Together, this validated epitope set achieved population coverage in ∼68% of patients, supporting potential utility in glioma vaccine design.
Conclusions
Most tools miss tumour-presented peptides in glioma, risking ineffective vaccine design. Our benchmarking identifies a recurrent, immunogenic, and manufacturable epitope set with broad coverage. Glioma-specific, MS-informed approaches are needed to improve neoantigen selection and accelerate vaccine development.
| Original language | English |
|---|---|
| Pages (from-to) | S487-S487 |
| Number of pages | 1 |
| Journal | Annals of Oncology |
| Volume | 36 |
| Issue number | Supplement 2 |
| DOIs | |
| Publication status | Published - 5 Nov 2025 |
| Event | European Society for Medical Oncology Congress 2025 - Berlin, Germany Duration: 17 Oct 2025 → 21 Oct 2025 https://www.esmo.org/meeting-calendar/esmo-congress-2025 |
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 3 Good Health and Well-being
Fingerprint
Dive into the research topics of '669P Mapping the neoantigen landscape of glioma in the UK brain cancer vaccine project'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver