Background: Aberrant DNA methylation is associated with the development of hepatocellular carcinoma (HCC), suggesting that gene methylation could be a potential biomarker for detection of HCC. The aim of this study is to identify potential biomarkers in HCC. Methods: We used the Infinium methylation array and a DNA-pooling strategy to analyze the genome-wide methylation profile in HCC. Quantitative methylation-specific PCR (Q-MSP) was used to validate homeobox A9 ( HOXA9) methylation in 29 normal controls, 100 HCC samples and adjacent non-tumor tissues and in 74 plasma samples, including 40 patients with HCC. Results: Ten genes ( HOXA9, NEUROG 1, TNFRSF10C, IRAK3, GFPT2, ZNF177, DPYSL4, ELOVL4, FSD1 , and CACNA1G ) showed differences in methylation between controls and HCCs. Of these, HOXA9 was significantly hypermethylated in HCCs (76.7%; 23/30) compared with controls (3.4%; 1/29). In addition, combination analysis of two- and three-gene sets for HCC detection showed greater sensitivity (90%–96.7%) and comparable specificity (93.1%–96.6%) to each individual gene (33.3%–76.7% and 55.2%–100.0%). HOXA9 methylation was further validated by Q-MSP in two independent set of clinical samples including 100 HCC and paired non-tumor tissues. Further, HOXA9 methylation could be detected in plasma from HCC patients (n=40) but not in normal plasma (n=34) (p<0.0005). Combined testing (either parameter positive) for α-fetoprotein (AFP, a plasma protein biomarker) and HOXA9 methylation showed greater sensitivity (94.6%) for detection of HCC than AFP alone (75.7%). Conclusions: These data suggest that methylation of HOXA9 could be a helpful biomarker to assist in HCC detection.