Résumé : High-resolution array comparative genomic hybridization (arrayCGH) profiling was performed on 75 primary tumors and 29 cell lines to gain further insight into the genetic heterogeneity of neuroblastoma and to refine genomic subclassification. Using a novel data-mining strategy, three major and two minor genomic subclasses were delineated. Eighty-three percent of tumors could be assigned to the three major genomic subclasses, corresponding to the three known clinically and biologically relevant subsets in neuroblastoma. The remaining subclasses represented (I) tumors with no/few copy number alterations or an atypical pattern of aberrations and (2) tumors with 11q13 amplification. Inspection of individual arrayCGH profiles showed that recurrent genomic imbalances were not exclusively associated with a specific subclass. Of particular notice were tumors with numerical imbalances typically observed in subtype I neuroblastoma, in association with genomic features of subtype 2A or 2B. A search for prognostically relevant genomic alterations disclosed 1q gain as a predictive marker for therapy failure within the group of subtype 2A and 2B tumors. In cell lines, a high incidence of 6q loss was observed, with a 3.87-5.32 Mb region of common loss within 6q25.1-6q25.2. Our study clearly illustrates the importance of genomic profiling in relation to tumor behavior in neuroblastoma. We propose that genome-wide assessment of copy number alterations should ideally be included in the genetic workup of neuroblastoma. Further multicentric studies on large tumor series are warranted in order to improve therapeutic stratification in conjunction with other features such as age at diagnosis, tumor stage, and gene expression signatures. © 2007 Wiley-Liss, Inc.