Hepatocellular carcinoma (HCC) most often develops in patients infected with hepatitis B or hepatitis C virus. Differential gene expression profiling is useful for investigating genes associated with lymph node metastasis (LNM). We screened genes to identify potential biomarkers for LNM in HCC. RNA was extracted from formalin-fixed specimens of paired intratumoral and peritumoral tissues of patients with lymph node-positive (n = 36) or negative (n = 36) HCC. A cDNA-mediated annealing, selection, extension, and ligation assay was done with an array of 502 known cancer-related genes to identify differentially expressed genes in 20 pairs of patients with or without LNM. Candidate biomarkers were evaluated by using immunohistochemistry and tissue microarrays in an independent cohort of 309 HCC patients who had undergone hepatectomy. Of the 309 patients, 235 (76.1%) patients were infected with hepatitis B. Compared with lymph node-negative patients, lymph node-positive patients had 17 overexpressed genes and 19 underexpressed genes in intratumoral tissues, and 25 overexpressed genes and 22 underexpressed genes in peritumoral tissues. Hypoxia-inducible factor (HIF)-1α, VEGF, and matrix metalloproteinase (MMP)-2 were selected for analysis in the cohort of 309 HCC patients. We found that intratumoral protein levels of HIF-1α, VEGF, and MMP-2 were independent risk factors for developing LNM. We identified 83 cancer genes that were differentially expressed in lymph node-positive and lymph node-negative HCC. Our findings show that the combination of intratumoral HIF-1α, VEGF, and MMP-2 may be useful as a molecular prediction model for LNM.