One of the great success stories of clinical oncology is the improvement in the cure rates of pediatric acute lymphoblastic leukemia (ALL) from around 10% in the 1960s to nearly 90% today. The primary factor responsible for this remarkable improvement is the personalization of treatment, with stratification of patients based on both disease and host characteristics in order to optimize therapy. While age, WBC, and immunophenotype provide a rudimentary system for classification of ALL, molecular factors are playing an increasingly important role in further individualization of ALL therapy. Such risk-based stratification strategies are also increasingly being used in the treatment of children with solid tumors. In addition, genomic technologies are now being used to identify new molecular markers or signatures for both diagnostic and prognostic purposes. Recently we reported the analysis of pediatric osteosarcoma by expression profiling in an attempt to identify a molecular signature that could predict the chemoresistance of a tumor before treatment is initiated. We identified a 45-gene signature that discriminates between good and poor responders to chemotherapy in osteosarcoma. Using this classifier, we can predict with 100% accuracy the chemoresponse of osteosarcoma patients prior to the initiation of treatment. These encouraging results suggest that the genomic approach will revolutionize the diagnosis and prognosis of pediatric cancer patients and improve their outcome through predictive, personalized care.