The Immune Barrier Atlas: Cancer-Type-Specific Resistance Mechanisms Across 33 Tumor Types and Their Matched Drug Classes

Immune checkpoint inhibitors fail in most patients, yet the dominant mechanism of resistance differs by cancer type. Using CIBERSORTx immune cell fractions from 11,373 tumors across 33 TCGA cancer types, we classified each sample by its primary immune barrier from four structurally derived categories corresponding to the four stages of the anti-tumor immune cycle: activation (effector T cells not primed or mobilised against the tumor), detection (insufficient antigen presentation to direct the response), drain (inflammatory myeloid cells overwhelming the effector response), and suppression (regulatory T cells and M2 macrophages actively inhibiting effectors). The classification reveals that 13 cancer types are suppression-dominant, 8 detection-dominant, 8 activation-dominant, and 4 drain-dominant. Each barrier maps to a specific drug class: CTLA-4 + PD-1 or IL-2/IL-15 for activation, cancer vaccines or STING agonists for detection, anti-VEGF + ICI for drain, and anti-CCR8 or Treg depletion for suppression. Cross-validation against four independent single-cell RNA-seq datasets (NSCLC n=242, melanoma n=37, breast cancer n=29, basal cell carcinoma n=11) confirms the classification, with each dataset identifying a different dominant barrier — one per stage of the cycle. The NSCLC scRNA-seq classification independently converges on Suppression at 42%, matching the TCGA bulk estimate of 41.8%. A consistent cross-cancer finding is that T cell exhaustion marks immune engagement rather than failure: exhausted T cell fractions are higher in responders than non-responders across all four scRNA-seq datasets (breast cancer 11-fold, p = 0.0003). The atlas provides a resource for matching drug class to resistance mechanism at the cancer-type level.