Integrative Bayesian Analysis of High-Dimensional Multi-Platform Genomics Data

• Activity Number: 647
• Type: Topic Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: International Indian Statistical Association
• Author(s): Kim-Anh Do*+ and Wenting Wang and Veera Baladandayuthapani and Bradley M Broom and Jeffrey S Morris and Ganiraju C Manyam
• Companies: MD Anderson Cancer Center and MD Anderson Cancer Center and and MD Anderson Cancer Center and MD Anderson Cancer Center and MD Anderson Cancer Center
• Keywords: Integrative ; Bayesian ; genomics ; High-throughput ; Sparsity priors ; glioblastoma

Abstract

Analyzing data from multiple platforms of genomics data with patients clinical outcomes helps us to understand the complex underlying biological process of a disease, as well as how this process relates to the prognosis of the disease. Our proposed model, integrative Bayesian analysis of genomics (iBAG), uses a hierarchical modeling technique to integrate multiple platforms as well as the biological regulatory networks underlying them into one model to discover genes important to patients clinical outcomes. With Bayesian sparsity priors, this framework adapts to the high-throughput data and performs coefficient estimation and variable selection simultaneously. Our simulation studies show that iBAG performs better (with a higher C-index) in discovering genes related to patients clinical outcomes compared to other existing integration models and the single-platform model. We applied iBAG to integrate gene expression and methylation data to study their associations with patient survival for the Cancer Genome Altas (TCGA) glioblastoma study. We discovered many new methylation-regulated genes that are potentially related to patient survival.