Abstract
Clear cell renal cell carcinoma (ccRCC) is a highly heterogeneous tumor that lacks reliable biological markers for diagnosis and prognostic monitoring. Currently, the differentially expressed genes between paired adjacent normal tissues and ccRCC tumor tissues at single-cell resolution remained to be further discovered. To address this challenge, we performed an integrative analysis of multiple single-cell databases containing paired ccRCC samples. Using the "CopyKAT" algorithm, we accurately identified ccRCC tumor cells. Subsequently, various pseudotime algorithms were employed to identify malignant cells with tumor stem cell-like properties and high plasticity. This cell subgroup exhibited high expression of malignant features, including hypoxia, epithelial-mesenchymal transition (EMT), and proliferation/invasion phenotypes. We then performed differential analysis to identify genes highly expressed in this subgroup and constructed a reliable clinical diagnostic model for ccRCC using multiple machine learning algorithms. Furthermore, we identified AXL as a key gene with significant oncogenic activity, where high expression of AXL correlated with poor patient prognosis. Immune infiltration and spatial transcriptomics analyses further revealed that AXL promotes tumor progression interaction with M2 macrophages. Taken together, our analysis establishes a reliable 13-gene panel diagnostic model and AXL gene as reliable biological markers for ccRCC, providing valuable targets and a theoretical foundation for the development of precision-targeted therapies for ccRCC.