Identification of an eight-gene signature for survival prediction for patients with hepatocellular carcinoma based on integrated bioinformatics analysis

Background Hepatocellular carcinoma (HCC) remains one of the leading causes of cancer-related death worldwide. Despite recent advances in imaging techniques and therapeutic intervention for HCC, the low overall 5-year survival rate of HCC patients remains unsatisfactory. This study aims to find a gene signature to predict clinical outcomes in HCC. Methods Bioinformatics analysis including Cox’s regression analysis, Kaplan-Meier (KM) and receiver operating characteristic curve (ROC) analysis and the random survival forest algorithm were performed to mine the expression profiles of 553 hepatocellular carcinoma (HCC) patients from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) public database. Results We selected a signature comprising eight protein-coding genes (DCAF13, FAM163A, GPR18, LRP10, PVRIG, S100A9, SGCB, and TNNI3K) in the training dataset (AUC = 0.77 at five years, n = 332). The signature stratified patients into high- and low-risk groups with significantly different survival in the training dataset (median 2.20 vs. 8.93 years, log-rank test P < 0.001) and in the test dataset (median 2.68 vs. 4.24 years, log-rank test P = 0.004, n = 221, GSE14520). Further multivariate Cox regression analysis showed that the signature was an independent prognostic factor for patients with HCC. Compared with TNM stage and another reported three-gene model, the signature displayed improved survival prediction power in entire dataset (AUC signature = 0.66 vs. AUC TNM = 0.64 vs. AUC gene model = 0.60, n = 553). Stratification analysis shows that it can be used as an auxiliary marker for many traditional staging models. Conclusions We constructed an eight-gene signature that can be a novel prognostic marker to predict the survival of HCC patients