Machine learning approach to predict prognosis and immunotherapy responses in colorectal cancer patients

The immune-related genes in the colorectal cancer (CRC) microenvironment are closely associated with patient prognosis and the efficacy of immunotherapy. In this study, a CRC risk model was established utilizing the expression profiles of immune-related genes. The risk prediction framework for CRC was created by integrating clinical and transcriptomic data through machine learning techniques. We incorporated 13 core immune-related genes ( IL18BP , RSAD2 , G0S2 , SIGLEC1 , SFRP2 , IFI44L , ISG20 , IFIT1 , OLR1 , SAMHD1 , HK3 , PTAFR , and CSF1 ), constructed a prognostic model and established Immune Response-related Risk Score (IRRS) model in CRC. IRRS strongly correlated with cancer staging, immune cell infiltration, immune cell activation, and the expression of genes associated with immunotherapy targets. Furthermore, this IRRS model outperformed the Tumor Immune Dysfunction and Exclusion (TIDE) tool in predicting immunotherapy response. Therefore, by integrating patient clinical and transcriptomic data and applying machine learning algorithms, we developed a predictive model with enhanced accuracy and clinical utility for risk stratification and immunotherapy response prediction in CRC patients. • Immune Response-related Risk Score (IRRS) outperforms Tumor Immune Dysfunction and Exclusion (TIDE) in predicting immunotherapy response. • IRRS enables precise and clinically relevant patient stratification. • Clinical and experimental validation confirm the IRRS as a reliable tool for personalized colorectal cancer treatment.