Abstract
PURPOSE: The potential role of epithelium-specific genes through the adenoma-carcinoma sequence in the development of colorectal cancer (CRC) remains unknown. Therefore, we integrated single-cell RNA sequencing and bulk RNA sequencing data to select diagnosis and prognosis biomarkers for CRC. METHODS: The CRC scRNA-seq dataset was used to describe the cellular landscape of normal intestinal mucosa, adenoma and CRC and to further select epithelium-specific clusters. Differentially expressed genes (DEGs) of epithelium-specific clusters were identified between intestinal lesion and normal mucosa in the scRNA-seq data throughout the adenoma-carcinoma sequence. Diagnostic biomarkers and prognostic biomarker (the risk score) for CRC were selected in the bulk RNA-seq dataset based on DEGs shared by the adenoma epithelium-specific cluster and the CRC epithelium-specific cluster (shared-DEGs). RESULTS: Among the 1063 shared-DEGs, we selected 38 gene expression biomarkers and 3 methylation biomarkers that had promising diagnostic power in plasma. Multivariate Cox regression identified 174 shared-DEGs as prognostic genes for CRC. We combined 1000 times LASSO-Cox regression and two-way stepwise regression to select 10 prognostic shared-DEGs to construct the risk score in the CRC meta-dataset. In the external validation dataset, the 1- and 5-year AUCs of the risk score were higher than those of stage, the pyroptosis-related genes (PRG) score and the cuproptosis-related genes (CRG) score. In addition, the risk score was closely associated with the immune infiltration of CRC. CONCLUSION: The combined analysis of the scRNA-seq dataset and the bulk RNA-seq dataset in this study provides reliable biomarkers for the diagnosis and prognosis of CRC.