IMIREG: a lineage-resolved regulon signature unveiling immune engagement archetypes and predicting immunotherapy response across diverse cancers.

A critical barrier to maximizing immune checkpoint blockade (ICB) efficacy is the lack of biomarkers that capture the core regulatory dynamics of the tumor microenvironment. We developed IMIREG, a 14-regulon transcriptional signature derived from inferred transcription factor activity that quantifies immune engagement. Validated across 50 immunotherapy cohorts (52 treatment arms) spanning 16 cancer types, IMIREG robustly predicts clinical benefit (mean AUROC = 0.71), consistently outperforming established response and resistance signatures, including the T cell-inflamed, IFN-γ, CD8 T effector, and antigen presentation signatures. Crucially, its predictive utility is specific to the immunotherapy context rather than general prognosis. Single-cell analysis of 21 datasets revealed selective IMIREG enrichment in checkpoint-restrained exhausted/memory/effector T cells and pro-inflammatory M1 macrophages, defining an "immune-engaged but restrained" state primed for therapeutic reactivation. Our lineage-resolved framework classifies IMIREG-high tumors into T cell-, macrophage-, and dual-driven archetypes, revealing that the dual-driven phenotype is markedly diminished in metastatic compared to primary tumors. In triple-negative breast cancer, longitudinal single-cell analysis of neoadjuvant anti-PD-1/radiotherapy biopsies establishes IMIREG as a dynamic pharmacodynamic marker that discriminates spatial response trajectories, and predicts pathological complete response in on-treatment biopsies collected prior to radiation initiation. Together, IMIREG provides a mechanism-based biomarker that captures anti-tumor immune regulatory circuitry, improving ICB patient stratification and illuminating immune dynamics in cancer progression.