Immune Evasion in Prostate Cancer: Resolving the Cold Tumour Paradox via a Hybrid Discrete–Continuum Computational Framework

Prostate cancer (PCa) is immunologically “cold” and resistant to immune checkpoint blockade (ICB), yet bulk analyses show low, non-prognostic PD-L1 expression. We hypothesised that this paradox reflects two overlooked dimensions: basal heterogeneity (static engine) and IFN-γ-driven adaptive resistance (adaptive engine). Using TCGA-PRAD data (n=554) to parameterise an agent-based model, we simulated clonal selection and extended it to a hybrid discrete-continuum framework with reaction-diffusion IFN-γ. Bulk PD-L1 was low (median 1.48 TPM) and non-prognostic (HR =1.15, p=0.621). The static engine alone produced weak immunoediting (1.10-fold enrichment), whereas the adaptive engine drove a 2.95-fold enrichment of PD-L1-high clones via protective sanctuary formation, without increasing final tumour burden. Induction knockout (Pmax=0) abrogated this advantage, while diffusion knockout (D=0) had no effect. The cold tumour paradox is resolved by a hierarchical twin engine: rare genomic outliers permit initial persistence, but local IFN-γ/PD-L1 feedback dominates resistance, identifying induction capacity as the primary therapeutic target for JAK/STAT inhibition combined with ICB.