Background The intricate crosstalk of various cell death forms was recently implicated in cancers, laying a foundation for exploring the association between cell death and cancers. Recent evidence has demonstrated that biological networks outperformsnapshot gene expression profiles at discovering promising biomarkers or heterogenous molecular subtypes across different cancer types. In order to investigate the behavioral patterns of cell death-related interaction perturbation in colorectal cancer (CRC), this study constructed the interaction-perturbation network consisting of 11 cell death pathways and further revealed four cell death network (CDN) derived subtypes (CDN1-4).
Methods We constructed a perturbation matrix of 11 cell death pathways by using colorectal cancer data from TCGA and GTEx, from which four CDNs were decoded and validated by three GEO datasets.
Results Four subtypes were well characterized and displayed distinct clinical and molecular features: (1) CDN1: elevated proliferative activity, frequent KRAS mutations, immune desert, and high tumor purity; (2) CDN2: stronger immune activation, best prognosis, high BRAF mutation frequency, high mutational burden, moderate proliferative activity, neoantigen burden, microsatellite instability, and might be sensitive to immunotherapy sensitivity; (3) CDN3: metabolically hyperactive, immune desert, and moderate prognosis; (4) CDN4: stroma-rich, worst prognosis, immune-suppressed advanced stage, strong tumor invasion, stem cell-like, high levels of EMT and TGF-β signaling, and AOC3 considered as a potentially predictive molecule for CDN4.
Conclusions In general, based on the construction of the cell death crosstalk network, which is more stable and effective than gene features, our study established four stable CRC molecular subtypes that could predict prognosis and guided treatment.