Accurate subclassification of leukemia and the identification of prognostic determinants are essential to guide therapy and to improve patients' outcome. According to present standards, pre-therapeutic assessment depends on a combination of different methods. We aimed to expand the molecular characterization of different acute leukemia subtypes to identify new genome-wide diagnostic markers. Total RNA from 90 adult patients suffering from acute lymphoblastic leukemia (ALL, n = 25) and acute myeloid leukemia (AML, n = 65) was extracted at diagnosis and high density oligonucleotide microarrays were used to analyze the expression profiles of 12,000/22,000 genes in all specimens (Affymetrix U95Av2/U133A). All cases were thoroughly characterized by individual combinations of cytomorphology, cytogenetics, multiparameter immunophenotyping, and molecular genetics. The expression signature of a small set of differentially expressed genes was sufficient to accurately discriminate eight clinically relevant acute leukemia subgroups. Underlying chromosomal aberrations or immunophenotypical characteristics were strictly correlated with a distinct gene expression pattern for AML with t(8;21), t(15;17), t(11q23)/MLL, or inv(16) as well as for precursor B-ALL with t(9;22), t(8;14), or t(11q23)/MLL and precursor T-ALL. These data support a possible future application of microarray technology for classification of the acute leukemias.