Integrated multi-omics and machine learning reveal an immunogenic cell death-related signature for prognostic stratification and therapeutic optimization in colorectal cancer.

Colorectal cancer (CRC) continues to rise in global incidence and remains a leading cause of cancer-related mortality. Immunogenic cell death (ICD) has emerged as a critical modulator of tumor microenvironment (TME) dynamics; however, its prognostic implications and therapeutic potential in CRC require systematic characterization. Through the integrative analysis of single-cell RNA sequencing and bulk transcriptomic data, 11 ICD-related genes with prognostic significance were identified in CRC. A comprehensive computational framework was then employed to evaluate 101 machine learning combinations, ultimately constructing an optimized 11-gene ICD-related signature (ICDRS) by integrating StepCox [forward] and RSF. The ICDRS exhibited strong predictive performance for overall survival in CRC patients across the training and validation datasets. Notably, the ICDRS-based nomogram achieved outstanding time-dependent AUCs (>0.90) for 1- to 3-year survival prediction. Multidimensional analysis revealed significant associations between ICDRS-derived risk score and distinct immune infiltration patterns, immunotherapy response and TME characteristics. Furthermore, a novel macrophage subtype, SPP1/SLC11A1, was discovered and characterized by high infiltration levels. Drug repurposing analysis indicated Olaparib as a potential therapeutic candidate for high-risk CRC patients. Therefore, this study establishes ICDRS as a promising tool for CRC prognosis and immunotherapy, with future validation studies planned to guide personalized treatment strategies.