Abstract: Objective The aim of this paper was to discover key genes that affect the prognosis of patients with hepatocellular carcinoma (HCC) by bioinformatics. Methods We searched the key genes in peripheral blood mononuclear cells (PBMCs) and tumor tissues in GEO and TCGA databases by using STRING to construct a protein-protein interaction (PPI) network, by using GEPIA to obtain the survival information, and by using Cytoscape to filter key gene modules for the analysis of differentially expressed genes (DEGs). The Cox proportional hazards regression model was applied for prognostic analysis. Results We obtained 225 DEGs in PBMC, of which 105 were up-regulated and 120 were down-regulated genes; out of which, six key genes were screened and verified by through key modules, e.g. GPSM2, PPIL1, POLR2H, CRNKL1, U2SURP and TRA2B; the high levels of these genes in tumor tissues was significantly related to the overall survival rate of patients; the clinical data of TCGA database were analyzed for prognosis, which showed that the GPSM2 level (HR=1.556,95%CI:1.153-2.100) was independently related to overall survival in patients with HCC stage III, and the trends of GPSM2, TRA2B and U2SURP in PBMCs were the same as in tumor tissues. Conclusion The high level of GPSM2 in HCC tumor tissues and PBMCs is independently associated with the poor prognosis of patients with HCC stage III, and it is expected this gene might become a new marker for screening and prognostic judgment of patients with HCC.

Key words: Hepatocellular carcinoma, Data retrieval, Key genes, Prognosis