Background
Although significant progress has been made in surgery, radiation therapy, chemotherapy, and targeted therapy, colorectal cancer (CRC) is still one of the main types of cancer in terms of global morbidity and cancer-related deaths. All the time, the TNM (tumor-nodes-matastasis) staging system have been used as three prognostic indicators of the risk of recurrence in CRC patients. But the TNM staging system only considers the anatomical characteristics of the tumor, not the biological characteristics of the tumor. The metabolic recoding of tumor cells helps them adapt to the tumor microenvironment. The tumor microenvironment can provide the energy needed to maintain the growth of their malignant tumor cells, including accelerating proliferation, anti-apoptosis, evading immune attack and maintaining cancer stem cell status[1]. Certain genetic drivers of CRC, such as p53 [2]and KRas[3], are well-known regulators of cancer metabolism. And metabolic gene variants promote colorectal cancer[4]. It is currently known that a single gene or molecular marker cannot provide a good diagnosis or predict the progression of the disease. A single biomolecular marker is usually unable to predict the survival of patients with COAD, and more and more research institutions are using multi-gene combination to build predictive models for disease diagnosis. TCGA and GEO provide a lot of tumor-related information, such as gene expression, methylation, mutations and clinical parameters[5], which are of clinical significance cancer biology has created unprecedented opportunities. In this study, we first screened the PRMG through univariate COX regression based on the expression of metabolic genes, and then used the LASSO to construct an important gene prognostic model. In addition, ROC curve analysis of independent prognostic factors was performed and a nomogram for predicting overall survival was constructed. GSEA shows the way of KEGG enrichment.