Tumor heterogeneity affects the precision of microarray analysis

Microarray-based analysis of global gene expression patterns defines groups of genes that correlate with specific tumor types and prognosis, but the identified genes may not all be of equal clinical utility due to technical factors that affect the precision of their measurement. To analyze how technical variability in measured expression levels may impact microarray-based analysis in a clinical setting, we used Ewing sarcoma/peripheral neuroectodermal tumor (EWS/PNET) in a model system that replicates the clinical scenario in which microarray-based analysis of gene expression will likely occur, namely analysis of a fresh tumor sample by a single chip. By comparing variability of measured expression due to purely technical factors with variability due to biologic factors, we confirm that variability is dependent on the level of gene expression. We also demonstrate that the variability in expression level from either cell line or tumor samples is significantly higher than can be attributed to specific probe sets that have an intrinsically poor performance. These results have significant impact on the application of cDNA microarray chip for molecular analysis performed in a clinical setting.