Renal cell carcinoma (RCC) is one of the most prevalent types of urological cancer. Exosomes are vesicles derived from cells and have been found to promote the development of RCC, but the potential biomarker and molecular mechanism of exosomes on RCC remain ambiguous. Here, we first screened differentially expressed exosome-related genes (ERGs) by analyzing The Cancer Genome Atlas (TCGA) database and exoRBase 2.0 database. We then determined prognosis-related ERGs (PRERGs) by univariate Cox regression analysis. Gene Dependency Score (gDS), target development level, and pathway correlation analysis were utilized to examine the importance of PRERGs. Machine learning and lasso-cox regression were utilized to screen and construct a 5-gene risk model. The risk model showed high predictive accuracy for the prognosis of patients and proved to be an independent prognostic factor in three RCC datasets, including TCGA-KIRC, E-MTAB-1980, and TCGA-KIRP datasets. Patients with high-risk scores showed worse outcomes in different clinical subgroups, revealing that the risk score is robust. In addition, we found that immune-related pathways are highly enriched in the high-risk group. Activities of immune cells were distinct in high-/low-risk groups. In independent immune therapeutic cohorts, high-risk patients show worse immune therapy responses. In summary, we identified several exosome-derived genes that might play essential roles in RCC and constructed a 5-gene risk signature to predict the prognosis of RCC and immune therapy response.
Keywords: RCC; clinical model; exosome; machine learning; prognosis.