Phase-resolved transcriptomic bottlenecks in peptide cancer vaccine response

Abstract

Background: Peptide cancer vaccines can elicit antigen-specific immunity, but peripheral immunogenicity often does not translate into durable tumor control. Methods: We analyzed transcriptomic profiles across three public human peptide-vaccine cohorts: C1/GSE278476, a MUC1 plus Poly-ICLC PBMC RNA-seq cohort with ordered anti-MUC1 IgG response classes; C2/GSE85698, manufactured dendritic-cell vaccine preparations linked to TARP ELISpot response; and C3/GSE53922, baseline PBMC expression linked to overall survival after personalized peptide vaccination in castration-resistant prostate cancer. Prespecified gene modules were summarized as mean standardized scores and tested with endpoint-appropriate cohort-level models with within-family FDR control. Results: Baseline immune-readiness was favorable in C1 (beta=0.301, p=0.0072, q=0.093; permutation p=0.0088) and associated with longer survival in C3 (HR=0.662, 95% CI 0.532-0.823, p=0.000206, q=0.00126). Baseline erythroid/inflammatory drag showed the opposite direction in C1 (beta=-0.258, p=0.031, q=0.202; permutation p=0.0324) and was associated with inferior survival in C3 (HR=1.390, 95% CI 1.181-1.636, p=0.0000755, q=0.000982). In C2, lower tolerogenic/myeloid dendritic-cell product-state expression was observed in strong ELISpot responders (8/19 focused genes q<0.05). At C1 week 2, a priming/costimulation/mTOR-AKT module showed an FDR-significant cross-sectional association with response class (permutation p=0.0026), but paired within-person change was not significant. Conclusions: Public peptide-vaccine transcriptomic data support a phase-linked model in which host readiness, erythroid/inflammatory drag, dendritic-cell product state, and early priming are measurable response-linked layers. These retrospective cohorts do not establish causality, biomarker status, clinical utility, or durable tumor control.