The increasing availability and maturity of DNA microarray technology has led to an explosion of cancer profiling studies. To extract maximum value from the accumulating mass of publicly available cancer gene expression data, methods are needed to evaluate, integrate, and intervalidate multiple datasets. Here we demonstrate a statistical model for performing meta-analysis of independent microarray datasets. Implementation of this model revealed that four prostate cancer gene expression datasets shared significantly similar results, independent of the method and technology used (i.e., spotted cDNA versus oligonucleotide). This interstudy cross-validation approach generated a cohort of genes that were consistently and significantly dysregulated in prostate cancer. Bioinformatic investigation of these genes revealed a synchronous network of transcriptional regulation in the polyamine and purine biosynthesis pathways. Beyond the specific implications for prostate cancer, this work establishes a much-needed model for the evaluation, cross-validation, and comparison of multiple cancer profiling studies.