Development of a machine learning-derived programmed cell death index for prognostic prediction and immune insights in colorectal cancer

Colorectal cancer (CRC) is a major contributor to cancer-related mortality worldwide, emphasizing the need for improved prognostic tools and therapeutic strategies. Programmed cell death, encompassing diverse modalities, plays a critical role in tumor biology and therapy response. Utilizing machine learning techniques, we developed a novel Programmed Cell Death Index (PCDI) incorporating multiple forms of PCD-related genes to predict outcomes in colorectal cancer CRC patients. The PCDI demonstrated robust prognostic performance, stratifying patients into high- and low-risk groups across multiple cohorts, with high PCDI scores correlating with poor survival, advanced tumor stage, and aggressive pathological features. A nomogram integrating PCDI with clinical variables showed strong predictive accuracy for 1-, 3-, and 5 year survival rates. Functional analysis revealed significant metabolic differences between high- and low-PCDI groups. Immune profiling identified associations between PCDI and immunosuppressive microenvironments, including elevated regulatory T cell levels and reduced PD-L1 expression in high-PCDI patients. Patients with high PCDI exhibited a potential resistance to immune checkpoint inhibitors. These findings emphasize PCDI's potential as a prognostic biomarker and a tool for guiding personalized therapeutic strategies in CRC patients.