Tumor-specific membrane-associated and secreted proteins represent a clinically important class of proteins because they have the potential to serve as serum diagnostic markers or immunotherapy targets. We have recently shown that, in addition to allowing gene expression profiling of human tumors, DNA microarrays can be used to identify new membrane-associated and secreted proteins. This can be accomplished by separating messenger RNA species bound to membrane-associated polysomes from other mRNAs and then quantifying the distribution of transcripts in the two fractions by hybridization to DNA microarrays1. Using this technique we have now identified approximately 4,000 human gene products that are likely to encode previously unrecognized secreted or membrane proteins. By combining this data set with gene expression profiling data sets from a variety of human malignancies (including those of the breast, liver, kidney, prostate, ovaries, brain and immune system) we have further identified a subset of these 4,000 genes that are candidates for the development of new serum diagnostics and immunotherapies. Many of these markers show more robust tumor-specific gene expression profiles than those currently used at the bedside. We present a paradigm for assigning priority to gene expression data for human malignancies and selecting candidates for downstream analysis that have the highest likelihood of achieving clinical success.