Pancreatic ductal carcinoma (PDC) is one of the most intractable human malignancies. Surgical resection of PDC at curable stages is hampered by a lack of sensitive and reliable detection methods. Given that DNA microarray analysis allows the expression of thousands of genes to be monitored simultaneously, it offers a potentially suitable approach to the identification of molecular markers for the clinical diagnosis of PDC. However, a simple comparison between the transcriptomes of normal and cancerous pancreatic tissue is likely to yield misleading pseudopositive data that reflect mainly the different cellular compositions of the specimens. Indeed, a microarray comparison of normal and cancerous tissue identified the INSULIN gene as one of the genes whose expression was most specific to normal tissue. To eliminate such a "population-shift" effect, the pancreatic ductal epithelial cells were purified by MUC1-based affinity chromatography from pancreatic juice isolated from both healthy individuals and PDC patients. Analysis of these background-matched samples with DNA microarrays representing 3456 human genes resulted in the identification of candidate genes for PDC-specific markers, including those for AC133 and carcinoembryonic antigen-related cell adhesion molecule 7 (CEACAM7). Specific expression of these genes in the ductal cells of the patients with PDC was confirmed by quantitative real-time polymerase chain reaction analysis. Microarray analysis with purified pancreatic ductal cells has thus provided a basis for the development of a sensitive method for the detection of PDC that relies on pancreatic juice, which is routinely obtained in the clinical setting.