Identification of Energy Metabolism Genes for the Prediction of Survival in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) samples were clustered into three energy metabolism-related molecular subtypes (C1, C2, and C3) with significantly different prognoses. Gene expression data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) were used. HCC energy metabolism-related molecular subtype analysis was conducted based on the 594 energy metabolism genes. Differential expression analysis yielded 576 differentially expressed genes (DEGs) among the three subtypes, which were closely related to HCC development. Six genes were finally selected from the 576 DEGs through LASSO-COX regression and used in constructing a six-gene signature-associated prognostic risk model, which was validated using the TCGA internal and GEO external validation cohorts. The risk model showed that high ANLN, ENTPD2, TRIP13, PLAC8, and G6PD expression levels were associated with high risk as the risk factors of prognosis and high expression of ADH1C was associated with low risk as the protective factor. The validation results showed that the model is robust, with areas under the curve (AUCs) of 0.66 and 0.65 and a good performance in prognostic prediction. The four enriched pathways of the risk model were obtained by gene set enrichment analysis (GSEA) and found to be associated with the occurrence and development of HCC, including the cell cycle, Wnt signaling pathway, drug metabolism cytochrome P450, and primary bile acid biosynthesis. The risk score calculated from the established risk model in 342 samples from the TCGA dataset and other clinical characteristics was used in building a nomogram with good prognosis and prediction of survival. The AUCs of the nomogram model in 1, 2, and 3 years were 0.82, 0.77, and 0.80, respectively. Then, qRT-PCR and immunohistochemistry were used to validate the mRNA expression levels of the six genes screened above, and significant differences in mRNA and gene expression were observed among the tumor and adjacent tissues. Overall, our study divided HCC patients into three energy metabolism-related molecular subtypes with significantly different prognosis. Then, a risk model with a good performance in prognostic prediction was built using the TCGA dataset. This model can be used as an independent prognostic evaluation index for HCC patients.