Ependymomas account for 6-12% of all pediatric intracranial tumors. Despite complete resection and radiation, about 50% of patients relapse and have subsequent dismal prognoses. As no clinical findings reliably forecast tumor recurrence, we sought to determine if gene expression profiling could be used to distinguish patients at high risk for relapse at initial diagnosis, and thereby make them candidates for innovative treatments at an early stage. We extracted RNA from 13 ependymoma specimens: 7 from patients who experienced tumor recurrence, and 6 from patients who have not recurred. RNA was applied to Affymetrix HG-U133 plus 2.0 microarray chips, and microarrays were analyzed with GeneSpring 7.0 and Prediction Analysis of Microarrays (PAM) software. The 3-gene subset of PLEK (pleckstrin), NF-kappaB2 (nuclear factor kappa beta-2), and LOC374491 (TPTE and PTEN homologous inositol phosphatase pseudogene) was identified as the minimal subset capable of accurately distinguishing tumors according to recurrence. In summary, gene expression profiling may be valuable, perhaps in combination with clinical findings identified in some studies, for identifying children at high risk for ependymoma relapse.