Childhood acute lymphoblastic leukemia (ALL) is a heterogeneous disease with several distinct genetic subtypes characterized previously by typical changes in gene expression pattern. Here, we analyzed microarray data of 31 ALL patients to identify changes in gene expression that are associated with current risk assignment, irrespective of the genetic subtype. Using all low- and high-risk patients, no gene was capable of predicting the risk assignment already at time of diagnosis. However, screening for risk group associated genes using more homogenous patient subsets revealed 106 discriminatory probe sets. Identified genes are involved in proliferation, cell cycle control, and apoptosis. Subsequently, the prognostic significance of these probe sets was determined for the entire patient collective. Using selected subgroups as training set and the remaining samples as independent test set, logistic regression using 3 predictor variables (the first principal component of 6 ribosomal proteins, as well as cyclin H and leucine-rich and death domain containing protein (LRDD/Pidd)) could accurately predict current risk assignment for 10 out of 12 patients. This suggests that gene expression data can support current parameters to predict risk classes of childhood leukemia already at time of diagnosis.