Pan-cancer Single Cell Atlas Reveals Tumor Hallmark Gene Modules and Their Clinical Relevance

Pan-cancer intratumour heterogeneity (ITH) characteristics and their relationship with clinical prognosis have not been fully investigated. In this study, we integrated and analyzed single-cell RNA sequencing(scRNA-seq) data from 106 samples of four tumor types, BRCA, HNSC, LUAD, and PRAD, identified 34 co-regulated genes and constructed a tumor hallmark gene set. The machine learning model results show that the tumor hallmark exhibits superior tumor sample identification. Tumor epithelial cells were classified into High and Low expression groups based on the scores of tumor hallmark gene set. Module analyses on the high-expression group using the hdWGCNA algorithm identified 13 distinct gene modules. Further pathway enrichment and spatial transcriptomics (ST) data analysis revealed four core modules with differential biological functions and spatial expression patterns. Especially, the green module was significantly enriched for vascular smooth muscle contraction and cytoskeleton in muscle cells pathways, and showed a significant positive correlation with M2 macrophages. The green module was also involved in the regulation of tumor angiogenesis and cellular structure, and may promotes tumor proliferation. High expression of green module genes was significantly associated with poorer clinical survival and tended to increase with clinical stage. Our study established a pan-cancer hallmark system and characterized key ITH modules, which provide an important annotation approach for understanding tumor heterogeneity and its clinical correlations.