Background The incidence of RCC has drastically increased in recent years. The large intratumor heterogeneity of RCC, especially ccRCC, usually leads to treatment failure. In addition, single biomarkers have a limited ability to predict prognosis. Therefore, we performed this study to select variables and provided a simple but efficient way to predict prognosis. Method Three studies from the GEO database were involved in the selection of DEGs. A total of 840 RCC patients and 524 ccRCC patients from the TCGA database were involved in the prognostic analyses. Nomograms based on the Cox regression model were used to select variables to predict the prognosis, and GSEA was used to demonstrate the potential pathways altered by gene expression. Result Our study suggested that DEGs existed between metastatic and primary tumor tissues. Loss of GLS2 function was related to poor prognosis in RCC and ccRCC. These results revealed that GLS2 expression combined with basic characteristics, including age and TNM stage, could efficiently predict prognosis. GLS2 serves as a tumor suppressor in ccRCC, and loss of GLS2 function endows cells with oncogenic functions and is related to advanced disease. According to the GSEA results, loss of GLS2 function may alter the cell cycle by activating the E2F pathway. Conclusion GLS2 is a tumor suppressor in RCC. Loss of GLS2 function in ccRCC predicts a poor prognosis via the E2F pathway. Nomograms based on DEGs and clinical features provide doctors with a simple but efficient way to predict prognosis. Further studies are needed to verify the pathway in our study.