Studying the Context-Specificity of Network Structure

• Activity Number: 56
• Type: Invited
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: WNAR
• Author(s): Karl Rohe*+ and Juhee Cho and Sushmita Roy
• Companies: UW Madison Statistics Dept and UW Madison Statistics Dept and UW Madison Biostatistics and Medical Informatics
• Keywords: network ; spectral ; latent space ; context-specificity ; ENCODE ; DNase1

Abstract

Relationships and networks exist in a context that can often be represented by covariates on the actors or by several different types of relationships. For example, in part of the ENCODE project, researchers used DNase1 to measure 41 different regulatory networks among the same 475 transcription factors. The 41 different networks correspond to 41 different types of human tissue (from embryonic stem cells to lung fibroblast cells). We aim to understand the dependence among these 41 networks. We propose a spectral technique that illustrates the joint dependence. Then, we develop a novel, convex framework that can estimate the "signature subgraphs" that create the joint dependence. Time permitting, the talk will present preliminary asymptotic estimation results under a general latent space model.