The invention provides selections of genes expressed in a cancer cell that function to characterize such cancer, and methods of using the same for diagnosis and for targeting the therapy of selected cancers. In particular, methods are provided to classify cancers belonging to distinct diagnostic categories, which often present diagnostic dilemmas in clinical practice, such as the small round blue cell tumors (SRBCTs) of childhood, including neuroblastoma (NB), rhabdomyosarcoma RMS), Burkitt’s lymphoma (BL), and the Ewing family of tumors (EWS). More specifically, the invention is an application of Artificial Neural Networks (ANNs) for the diagnostic classification of cancers based on gene expression profiling data derived from cDNA microarrays. The ANNs were trained using as models. The ANNs then correctly classified all samples tested and identified the genes most relevant to the classification. Their study demonstrated the potential applications of these methods for tumor diagnosis and for the identification of candidate targets for therapy. The uniqueness of this method is taking gene expression data generated by microarrays, minimizing the genes from the original 1000s to less than 100, identifying which genes are the most relevant to a classification, which gives an immediate clue to the actual biological processes involved, not just surrogate markers which have no bearing on the biology.