Abstract
BACKGROUND: Hepatocellular carcinoma (HCC) is one of the most lethal cancers, with a poor prognosis. Prognostic biomarkers for HCC patients are urgently needed. We aimed to establish a nomogram prediction system that combines a gene signature to predict HCC prognosis. METHODS: Differentially expressed genes (DEGs) were identified from publicly available Gene Expression Omnibus (GEO) datasets. The Cancer Genome Atlas (TCGA) cohort and International Cancer Genomics Consortium (ICGC) cohort were regarded as the training cohort and testing cohort, respectively. First, univariate and multivariate Cox analyses and least absolute shrinkage and selection operator (LASSO) regression Cox analysis were performed to construct a predictive risk score signature. Furthermore, a nomogram system containing a risk score and other prognostic factors was developed. In addition, a correlation analysis of risk group and immune infiltration was performed. Finally, we validated the expression levels using real-time PCR. RESULTS: Ninety-five overlapping DEGs were identified from four GEO datasets, and we constructed a four-gene-based risk score predictive model (risk score = EZH2 * 0.075 + FLVCR1 * 0.086 + PTTG1 * 0.015 + TRIP13 * 0.020). Moreover, this signature was an independent prognostic factor. Next, the nomogram system containing risk score, sex and TNM stage indicated better predictive performance than independent prognostic factors alone. Moreover, this signature was significantly associated with immune cells, such as regulatory T cells, resting NK cells and M2 macrophages. Finally, RT‒PCR confirmed that the mRNA expressions of four genes were upregulated in most HCC cell lines. CONCLUSION: We developed and validated a nomogram system containing the four-gene risk score, sex, and TNM stage to predict prognosis.