Objective To identify genes associated with hepatocellular carcinoma (HCC) as candidate diagnostic markers in a genome-wide scale. Methods The gene expression profiles of 40 pairs of HCC tumor tissue and peripheral non-tumorous liver tissue were analyzed by using gene chip technology.The gene chips were fabricated at the National Cancer Institute (NCI). Each gene chip contained 9 180 genes. The fluorescent targets were prepared by a direct labeling approach using two kinds of fluorescences as following: 100 ?g of total RNA from non-cancerous liver tissue was labeled with Cy3-dUTP and 200 ?g of total RNA from HCC was labeled with Cy5-dUTP. The targets were mixed together and hybridized with genes on the gene chips. Unsupervised hierarchical clustering analysis was done by CLUSTER and TREEVIEW software using median centered correlation and complete linkage. Results A total of 10 genes were found up-regulated in over 80% of primary tumors comparing with that of their corresponding non-tumorous liver tissues at a two-fold filter with an unsupervised hierarchical clustering algorithm, including protocadherin-alpha 9, ESTs, Homo sapiens cDNA FLJ, KPNA2, RPS20, SNRPE, CDKN2A, UBD, MDK and ANXA2. Conclusion These genes are supposed to be candidates for the diagnosis of HCC. Further investigation of these genes in a large scale of patients with HCC and patients with non-malignant hepatic diseases will be needed to disclose whether they could be used clinically as novel diagnostic tumor markers for HCC.