Integration of Omics Data to Study Complex Phenotypes

• Activity Number: 58
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Katerina Kechris*+ and Daniel Dvorkin
• Companies: Colorado School of Public Health and Altitude Research Center
• Keywords: genomics ; hierarchical mixture model ; data integration

Abstract

Making effective use of multiple data sources is a major challenge in genomic research. Genome-wide data such as measures of transcription factor binding, gene expression, and sequence conservation can each provide valuable information for understanding the processes under study. However, these heterogeneous data types can be difficult to analyze together due to differences in biological meanings, genomic scale and statistical distributions. Here we present methods for integrating multiple data sources to identify genes that play specific biological roles. We describe a family of hierarchical mixture models and computationally efficient fitting procedures for data integration with clear biological and statistical interpretations. An un-supervised approach is presented, along with a semi-supervised extension when a small training set is available. Motivating examples include the identification of genes involved in developmental pathways in fly, essential genes in yeast, and oncogenes in human using data from The Cancer Genome Atlas.