An ICD gene set-derived immune contexture signature for colorectal cancer prognosis: integrated single-cell and bulk transcriptomic analysis with external validation.

Immunogenic cell death (ICD) links tumor cell demise with antitumor immunity, but the transcriptional features associated with ICD gene expression patterns and their prognostic significance in colorectal cancer (CRC) remain areas of active investigation. This study integrated single-cell and bulk transcriptomic data from TCGA and GEO to characterize ICD gene set-derived transcriptional features in CRC. ICD-associated gene modules were identified through weighted gene co-expression network analysis (WGCNA) independently in colon and rectal cancers. A seven-gene immune contexture signature (ICS) - CD79A, CXCR6, IRF4, ISG20, PLCG2, TIGIT, TRAF1 - was derived using random survival forest, gradient boosting machine, and Lasso-Cox regression. These genes are immune effector molecules rather than canonical ICD mediators (calreticulin, ATP, HMGB1); the signature should be interpreted as an ICD gene set-derived immune contexture score reflecting the immunological correlates of ICD-associated gene expression, not a direct measure of ICD induction. In the TCGA-CRC training cohort, the signature stratified patients (median cutoff: P = 0.001, HR = 1.906) with 1-, 2-, and 5-year AUCs of 0.68, 0.68, and 0.58, respectively. External validation in GSE39582 (n = 561) showed a non-significant trend (P = 0.072, HR = 1.298). Single-cell expression profiling confirmed that all seven genes were predominantly transcribed by immune cells. The risk-score effect was attenuated after adjustment for immune infiltration estimates. This study provides a hypothesis-generating ICD gene set-derived immune contexture framework, but the signature's modest predictive performance, non-significant primary external validation, and correlative nature indicate that independent validation is required before any clinical application.