The majority of ovarian tumors arise from the transformation of the ovarian surface epithelial cells, a single layer of cells surrounding the ovary. To identify genes that may contribute to the malignant phenotype of ovarian cancers, cDNA representational difference analysis was used to compare expressed genes in primary cultures of normal human ovarian surface epithelium (HOSE) and ovarian tumor-derived epithelial cells from the Cedars-Sinai Ovarian Cancer (CSOC) repository. A total of 255 differentially expressed genes were identified, of which 160 and 95 were specifically expressed in HOSE and CSOC cells, respectively. Using cDNA array hybridization, the expression profiles of the genes identified by cDNA-representational difference analysis were examined in an additional 5 HOSE and 10 CSOC lines. The comparison of average signal of each gene revealed 44 HOSE-specific and 16 CSOC-specific genes that exhibited at least a 2.5-fold difference in expression. A large number of genes identified in this study encode membrane-associated or secreted proteins and, hence, may be useful as targets in the development of serum-based diagnostic markers for ovarian cancer. Very few genes associated with protein synthesis or metabolism were identified in this study, reflecting the lack of observable differences in phenotypic or growth characteristics between HOSE and CSOC cells. Northern blot analysis on a subset of these genes demonstrated comparable levels of gene expression as observed in the cDNA array hybridization.