Tumor genesis and development often result from deregulation of important biological pathways at the gene expression level. Although there has been much work focused on searching gene sets using gene expression data or other prior information, proper statistical testing of the gene sets is still an open question. Most studies have expanded the testing method of a single gene into the gene sets. Parametric statistical analysis of gene sets ( p-SAGE ) was presented for determining the significant gene sets or pathways associated with a phenotype of interest. The method was applied to brain tumor experiments to identify many gene sets. Some of the newly discovered gene sets were related to signal transduction and immunity. This simple and effective method gives useful biologically meaningful results.