Prediction of Active Molecular Modules Through Integrated Expression Analysis

• Activity Number: 127
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Christine Duarte*+
• Companies: Maine Medical Center
• Keywords: genomics ; bioinformatics ; data mining ; gene expression ; high dimensional

Abstract

In silico discovery of perturbed molecular pathways via integrative analysis of high dimensional genomic data is a powerful approach for discovering molecular mechanisms of disease. However, data mining of genomic data sets at multiple levels simultaneously offers new challenges for traditional analytic approaches. To address the gap in methodology for joint analysis of cross-platform genomic data, we have developed a method for predicting dysregulated molecular pathway modules from the simultaneous analysis of microRNA expression, gene expression, and clinical phenotype data. The method is based on the formalism of Dynamic Bayesian Networks, with network module discovery accomplished using a newly proposed algorithm. We demonstrate this method using data for Acute Myeloid Leukemia from the Cancer Genome Atlas Project (TCGA) to predict active regulatory modules that influence the survival of AML patients. We show the discovery of two gene expression modules with predicted regulators that are members of the miR-181 family, a family of miRNA that is well-known to be involved in leukemia.