We use a targeted approach to discover biomarkers for assay development. First, markers for cancer detection and disease stratification are identified by comparing the transcriptomes, or gene expression, of cell types within tumors to those of their normal counterpart. Cell type-specific transcriptomes are determined via cell isolation from tissue specimens using antibodies to cell surface CD antigens followed by expression analysis using Affymetrix DNA arrays. Second, tumor upregulated genes encoding extracellular/secreted proteins are selected. The gene AGR2 showed a comparatively high level of expression in prostate cancer cells, and the 17-kDa AGR2 protein was detected in tumor tissue preparations by Western blot analysis using a commercial mouse antibody 1C3. Third, tissue microarray analysis validated prostate tumor expression of AGR2, and non-cancer tissue showed little AGR2 expression. Fourth, new monoclonal antibodies are being generated against AGR2 for the development of a urine test to screen men with prostate cancer, where tumor secreted AGR2 protein might be present in voided urine at ?ng/ml levels. The 1C3 antibody, which was raised against a bacterially produced recombinant AGR2 protein, appeared not to recognize native AGR2 as secreted by the AGR2-I- prostate cancer cell line CL1. A novel procedure to obtain useful antibodies is also proposed to immunize mice with proteins in tumor tissue preparations. Tumor heterogeneity could be attributed to multiple cancer cell types: Gleason pattern 3 vs. Gleason pattern 4; CD10- vs. CD10+. Differentially expressed genes among these cell types are useful as risk assessment markers because pattern 4 and CD10+ cancer cells are associated with poor outcomes. In addition, markers could be derived from differentially expressed genes of tumor-associated stromal cells. One example is the ~29 kDa CD90/THY1 protein found released into tumor tissue preparations. A multi-marker analysis tool, multiple reactions monitoring (MRM) mass spectrometry, is being developed to measure multiple informative proteins simultaneously in urine. Cell transcriptomes allow the development of cancer cell type- specific signatures that consist of genes and their abundance (in transcript counts). These can be applied to diagnose tumors for the presence of specific cancer cell types, which will then provide prognostic information.