Inferring Disease-Related Interaction Network

• Activity Number: 417
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: IMS
• Author(s): Qi Zhang*+ and Pei Wang
• Companies: Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
• Keywords: Network ; Partial Correlation ; Sparse Regression ; Differential Co-expression Analysis ; High Dimension Data

Abstract

Detecting disease-related interactions among genes or proteins are essential for discoverying disease responsible pathways. We propose an efficient method for constructing disease related interaction network based on high dimensional genomic array data. Specifically, the method assumes the overall sparsity of the network and infers interactions through non-zero partial correlations. It employs varying-coefficient models in a joint sparse regression framework to characterize the dependence of interactions on disease status. Compared with the existing approach which infers and then compares multiple networks fitted for different disease groups, the proposed new method has the advatange of exploiting the entire set of samples coherently, and can be naturally used for continuous disease phenotypes. We illustrate the performace of our method through simulation studies and experimental results