Chromosomal numerical aberrations (CNAs), particularly regional amplifications and deletions, are a hallmark of solid tumor genomes. These genomic alterations carry the potential to convey etiologic and clinical significance by virtue of their clonality within a tumor cell population, their distinctive patterns in relation to tumor staging, and their recurrence across different tumor types. In this study, we showed that array-based comparative genomic hybridization (CGH) analysis of genome-wide CNAs can classify tumors on the basis of differing etiologies and provide mechanistic insights to specific biological processes. In a RAS-induced p19(Arf-/-) mouse model that experienced accelerated melanoma formation after UV exposure, array-CGH analysis was effective in distinguishing phenotypically identical melanomas that differed solely by previous UV exposure. Moreover, classification by array-CGH identified key CNAs unique to each class, including amplification of cyclin-dependent kinase 6 in UV-treated cohort, a finding consistent with our recent report that UVB targets components of the p16(INK4a)-cyclin-dependent kinase-RB pathway in melanoma genesis (K. Kannan, et al., Proc. Natl. Acad. Sci. USA, 21: 2003). These results are the first to establish the utility of array-CGH as a means of etiology-based tumor classification in genetically defined cancer-prone models.