PURPOSE: Breast cancer is characterized by an extensive heterogeneity that complicates the precise assessment of tumor aggressiveness and makes therapeutic decisions difficult. Variation in transcriptional programs accounts for much of the biological diversity of the tumors. We tried to investigate the gene expression profile of Korean breast cancer with cDNA microarray technique. METHODS: We have characterized variation in gene expression patterns in 7 tissue samples of infiltrating ductal carcinoma from different individuals and 1 tissue sample of malignant phyllodes tumor using cDNA microarray representing 7,500 human genes. Breast cancer cell line, MCF7 and T lymphoblastoid cell line, CCRF-CEM were also included in this study. RNA xtracted from normal breast tissue of 3 non-cancer individuals were pooled and used as a control sample. We analyzed the data with Cluster and TreeView program. RESULTS: Tissues from infiltrating ductal carcinoma and tissue from phyllodes tumor and 2 celllines are clearly differ entiated by hierarchical clustering as their pathologic features and showed characteristic genetic expression. Some inter esting clusters were found. They were 'stromal cluster','immune cluster', 'proliferation and transcription cluster', and a cluster including FOS and JUNB gene. Phylldes tumor showed distinctive genetic expression pattern compared with ductal carcinoma tissues. Two patients whose c-erbB2 protein had been measured as highest level in immuno histochemistry showed overexpression of ERBB2 gene. GRB7, MLN51, and PPARBP in 17q21 that are known to coexpress with ERBB2 also showed overexpression in these patients. CONCLUSION: We found some interesting clusters and confirmed the feasibility of cDNA microarray in the study of breast cancer. This preliminary study will help the further research of DNA microarray with lager sample size.