Construction and validation of a metabolism-related prognostic model for thyroid cancer

Metabolic reprogramming is a common pathological process of cancer. Expression of metabolism-related genes differs in thyroid cancer (TC) patients with different prognoses. This work committed to constructing a prognostic model for TC through identifying metabolism-related signatures. Expression profiles of mRNAs and clinical data of TC, were acquired from The Cancer Genome Atlas. Differential analysis was performed on mRNA expression profiles. The obtained differentially expressed genes (DEGs) were overlapped with metabolism-related genes from MSigDB database to acquire metabolism-related DEGs. Cox regression and Least Absolute Shrinkage and Selection Operator analyses were performed to ascertain feature genes and to build a prognostic model for TC. The model was evaluated comprehensively through survival curve, time-dependent receiver operating characteristic (ROC) curve, gene set enrichment analysis (GSEA), and Cox regression analyses combining varying clinical information. 7 key genes related to metabolism, including AWAT2, GGT6, ENTPD1, PAPSS2, CYP26A, ACY3 and PLA2G10, were identified, based on which a prognostic model was constructed. The survival analysis indicated that high-risk group presented shorter survival time than low-risk group. ROC curve results exhibited that AUC values of 3-year and 5-year survival of TC patients were both >0.70. Besides, GSEA on high/low-risk groups revealed that DEGs were mainly gathered in biological functions and signaling pathways linked with keratan sulfate catabolism and triglyceride catabolism. Combined with clinical information, Cox regression analyses unveiled that the 7-gene prognostic model can be an independent predictor. In conclusion, this model can effectively predict prognoses of TC patients, and also offer guidance for clinical treatment of TC.