Genomic approaches are providing a wealth of information on differential gene expression in cancer. To identify the most interesting genes amongst the many identified, high-throughput methods for analysis of genes at the translational level are required. We have used a rapid method for the in vitro selection of antibodies to peptide antigens for the generation of probes to 5 gene products that we have found to be overexpressed in colorectal cancer. The rationale of our study was to select a non-immune phage displayed human antibody library on peptides designed from the coding regions of the gene sequences and to verify whether such antibodies would be suitable probes for the parental protein in immunohistochemical and Western blot analysis. After the generation of a profile of genes overexpressed in primary colorectal cancer (CRC) we selected 5 genes, Ese-3b, Fls353, PBEF, SPARC and Smad5 for a more detailed analysis using phage display-derived antibodies. For these 5 antigens we designed 14-20 amino acid peptides predicted to be exposed on the surface of the parental protein. Selection of a large phage displayed antibody library resulted in specific antibodies for 6 of 8 different peptides with between 2 and 15 different antibodies isolated per peptide. Of 20 antibodies tested, 2 antibodies recognized the putative parental protein from primary CRC tissue. An antibody specific for a PBEF-derived peptide (Fab/PBEF-D4) was shown to recognize a protein product of the expected molecular weight in Western blotting and showed overexpression in n = 6/8 matched tumor/normal protein lysates. Furthermore, in immunohistochemistry this antibody showed restricted staining of the tumor stromal compartment with no detectable staining of epithelial cells. The discovery that PBEF is overexpressed in cancer is unexpected given that the normal function of PBEF is as a cytokine required for the maturation of B cell precursors. We also report on the isolation of an antibody (Fab/SMAD-50) specific for a Smad5-derived peptide that showed cytoplasmic staining of epithelial cells in both CRC tumor and matched normal mucosa. Fab/SMAD-50 also bound to a group of proteins in Western blotting with molecular weights consistent with belonging to the Smad family. These antibodies may be suitable probes for further investigation of the roles of PBEF and Smad5 in cancer. The amenability of phage display to automation suggests that this approach may be developed for implementation on a genomics scale. Indeed, the large-scale generation of antibody probes that can be used to study protein expression in situ would be of great value in target validation for functional genomics.