Cross-platform classification in microarray-based leukemia diagnostics

Abstract

Gene expression profiling is a powerful technique for classifying hematologic malignancies. Its clinical use is, however, currently hindered by the need to collect large sets of expression profiles at each diagnostic facility. To overcome this limitation, we introduced cross-platform classification, allowing classifier construction using pre-existing microarray datasets. As proof-of-principle, we performed cross-platform classification of acute myeloid leukemia and childhood acute lymphoblastic leukemia using expression data from four different facilities. We show that cross-platform classification of these disorders is achievable, and, strikingly, that the diagnostic accuracy can be retained. We conclude that cross-platform classification constitutes an effective and convenient way to implement microarray diagnostics.