Neuroblastomas show remarkable biological heterogeneity, resulting in favorable or unfavorable outcomes. To survey the differences in gene expression profiles between favorable and unfavorable neuroblastomas, we analyzed ten favorable neuroblastoma samples from patients whose tumors consequently regressed or matured and ten unfavorable tumor samples from patients who consequently died of disease using the microarray technique. In each sample, total RNA was labeled with Cy3 or Cy5 in reverse-trancriptase reaction and hybridized with our original microarray prepared with a cDNA library of human fetal brain. Microarray analysis revealed that 43 genes, including MYCN, hTERT, NME1 and cell cycle regulatory protein-coding genes, were highly expressed in unfavorable neuroblastomas, while another 80 genes were detected as highly expressed in favorable tumors, including neuronal differentiating genes and apoptotic inducing genes. Among favorable neuroblastoma samples, highly expressing genes in regressing tumors were different from those in maturing tumors. Expression profiling data revealed the existence of up-regulated and down-regulated gene clusters in favorable and unfavorable tumors. This cluster analysis is a powerful procedure to distinguish unfavorable tumors from favorable tumors as well as regressing tumors from maturing tumors among favorable tumors. The information obtained from expression profiling would clarify the key genes for cell growth, regression or maturation of neuroblastoma cells, and these genes will become diagnostic and therapeutic targets in human neuroblastoma in the future.