Abstract: Objective To screen methylated CpG diagnostic biomarkers for hepatocellular carcinoma (HCC). Methods The DNA methylation and gene expression data of HCC were downloaded from the cancer genome atlas (TCGA) database for bioinformatics analysis to obtain the candidate CpG sites. The methylation rates of candidate CpG sites and the mRNA expression of corresponding genes in 50 early-stage HCC tissue samples were detected by pyrosequencing and qRT-PCR, to verify the diagnostic efficacy. The diagnostic efficacy of candidate CpG sites was evaluated in datasets GSE54503, GSE89852, and GSE56588, respectively. Results Five CpG sites(cg12614630, cg19786751, cg06131338, cg23371746, and cg25340966) were screened out by bioinformatics analysis. The area under curve (AUC) of receiver operating characteristic (ROC) curve of multivariable combined diagnosis was 0.993. The methylation rates of cg12614630 (GPR182), cg19786751 (ACACB), and cg06131338 (ACACB) in 50 early-stage HCC tissues were higher than those in adjacent tissues, while the corresponding gene expression levels were lower than those in adjacent tissues, and site methylation was negatively correlated with gene expression levels (P<0.05). ROC curve analysis found that in the early-stage HCC tissues, GSE54503, GSE89852 and GSE56588, the AUCs of the combined diagnosis of three CpG sites were 0.903, 0.812, 0.844 and 0.934, respectively. Conclusions The CpG sites cg12614630, cg19786751, and cg06131338 may be potential diagnostic biomarkers for HCC, and the combined diagnosis of the these CpG sites has favorable diagnostic performance.

Key words: HCC, Diagnostic biomarker, DNA methylation, CpG site