Discovering Multivariate Gene Expression Markers for Relapse Prediction in Pediatric Acute Lymphoblastic Leukemia
Y Zhou, D Dziuda. New York Medical College, Valhalla, NY; Central Connecticut State University, New Britain, CT
Background: Identification of cancer patients with high risk of treatment failure is becoming an important aspect of pathology practice. Discovery of reliable predictive and prognostic markers using high content gene expression profiling or genome analysis have proven both promising and challenging. Oncogenesis is a result of heterogeneous signaling dysregulation converged on critical cellular functions. The goal in discovery of predictive and prognostic marker is to identify a minimal set of genes or proteins that has maximal discriminating power to separate phenotypic classes, and reveals the underlying mechanism contributing to the phenotypic differences.
Design: We describe a supervised multivariate biomarker discovery approach utilizing stepwise selection driven by a well-defined measure of discriminatory power (the Lawley-Hotelling T 2 metric), which (i) allows for identification of a set of genes whose joint expression pattern efficiently discriminate phenotypic classes, (ii) facilitates biological interpretation of underlying dysregulations. This approach is illustrated on prediction of relapse in pediatric B-cell acute lymphoblastic leukemia (B-cell ALL) using publicly available gene expression data sets.
Results: Based on a cohort of B-cell ALL with mixed subtypes of cytogenetic changes, a set of genes, less than 20, was identified as a multivariate biomarker for predicting the risk of disease relapse. This marker is then tested on another independent set of B-cell ALL, and predicts disease relapse with 91% sensitivity. Ensemble of the biomarker reveals gene expression pattern that is associated with the relapse risk in B-cell ALL.
Conclusions: We demonstrate the usefulness of multivariate approach in identifying a small set of prognostic marker. The marker contains sufficiently small numbers of genes that can be readily converted to RT-PCR based assays.
Category: Pan-genomic/Pan-proteomic Approaches to Diseases
Monday, March 22, 2010 1:00 PM
Poster Session II # 235, Monday Afternoon