What this research found
Pancreatic ductal adenocarcinoma (PDAC) carries few mutations, which makes a personalized cancer vaccine a hard proposition. Genomic, structural, clinical, and tumor-microenvironment evidence was assembled to establish where individualized neoantigen-specific therapy (iNeST) can still gain traction: TCGA cohorts put median PDAC tumor mutational burden at 1.21 mutations per megabase against 7.61 for melanoma, and the epitope filtering funnel leaves a median of three immunogenic candidates per patient. The recommendation is quality-driven epitope selection around KRAS hotspots paired with stromal and checkpoint co-targeting.
- PDAC is mutationally quiet next to the tumors where immunotherapy works. Median tumor mutational burden was 1.21 mutations per megabase across 150 pancreatic adenocarcinomas, against 7.61 in 368 melanomas and 4.09 in 230 lung adenocarcinomas (Kruskal-Wallis H = 224.64, p = 1.66e-49; PDAC versus melanoma Bonferroni p = 5.77e-42).
- What mutations exist are concentrated in a handful of genes: KRAS in 136 of 150 samples (90.7%), TP53 in 104 (69.3%), SMAD4 in 37 (24.7%), and CDKN2A in 22 (14.7%). Every sample carried at least one mutation among the top 30 genes.
- The obvious target sits in an awkward place. Solvent-accessibility calculations on the wild-type KRAS structure put glycine-12 at 9.9 Ų of accessible surface and a relative accessibility of 0.102, partially buried at the nucleotide-binding interface and below the 0.20 exposure threshold, so G12D and G12V only become visible to T cells after proteasomal processing and MHC-I presentation.
- Epitope yield is thin. A median of 75 somatic mutations per patient narrows through coding (52), expressed (34), MHC-I binding (18), MHC-II binding (12), and high-confidence (7) filters to 3 immunogenic candidates — roughly 4% retention — which is the case for selecting epitopes on quality rather than counting them.
- Even good epitopes meet an immune-excluded tumor. Open Targets association scores rank TGFBR2 (0.3719), PD-L1/CD274 (0.3613), CXCR4 (0.2828), STAT3 (0.2808), and CTLA4 (0.2805) as the leading stromal and checkpoint targets, supporting combination regimens over vaccine monotherapy.
- The clinical signal behind the strategy is a 2023 Nature report in which patients mounting an iNeST T-cell response showed a hazard ratio for recurrence of 0.06 (p = 0.013). The Phase II IMcode003 trial (NCT05968326) is now recruiting 260 patients to test autogene cevumeran plus atezolizumab with disease-free survival as its endpoint.
How it was done
Mutation calls for three TCGA cohorts were pulled from cBioPortal and tumor mutational burden computed per sample at a 38 Mb exome size, then compared with Kruskal-Wallis and Bonferroni-corrected pairwise tests. The top 30 mutated PDAC genes were annotated against UniProt and scored for immunogenic potential by mutation type, domain overlap, and recurrence. Structures for wild-type KRAS, KRAS G12D, and an HLA-A MHC-I molecule were downloaded from the Protein Data Bank and per-residue solvent accessibility computed with the Shrake-Rupley algorithm to test whether the G12 site is surface exposed. Three registered neoantigen vaccine trials and 67 PubMed articles from 2021 to 2026 were then synthesized alongside 5,000 Open Targets associations for pancreatic ductal adenocarcinoma to rank stromal immunosuppression targets. The output is a 36-slide deck carrying nine data figures, including a mutation oncoplot, a 3D KRAS structure view, a Sankey diagram of the epitope funnel, and a tumor microenvironment pathway network.
Data sources
- cBioPortal TCGA Firehose Legacy cohorts — paad_tcga (150 sequenced samples), skcm_tcga (368), luad_tcga (230); 22,409 non-silent PDAC mutations
- UniProt — reviewed human entries for all 30 top-mutated genes, 29 with annotated domains
- RCSB Protein Data Bank — 4OBE (KRAS wild type), 4DSN (KRAS G12D), 1AO7 (HLA-A MHC-I)
- Open Targets Platform v4 — 5,000 of 5,165 associations for pancreatic ductal adenocarcinoma (MONDO_0005184)
- ClinicalTrials.gov — NCT05968326, NCT04161755, NCT03953235
- PubMed — 67 articles across four topics, 2021–2026
- Löffler et al., Nature (2023) — iNeST immune correlates of response
Limitations
Tumor mutational burden was derived from exome mutation calls with a fixed 38 Mb exome size, and the epitope funnel reports median per-patient figures rather than a patient-level prediction run. The clinical evidence supporting the strategy comes from small early-phase trials, one of which accrued only 29 patients.
Figures from this analysis
How this research was produced
K-Dense Web planned and ran this cancer biology investigation end to end — gathering the sources, carrying out the analysis, producing the figures, and drafting the report. The full session transcript, including every intermediate step, is available to view.
