A Discussion of Current Applications and Statistical Issues in Genomic Prediction of Complex Traits and Diseases

• Activity Number: 566
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Christine Woods Duarte*+ and Gustavo de los Campos
• Companies: University of Alabama at Birmingham and University of Alabama at Birmingham
• Keywords: complex traits ; GWAS ; prediction ; QTL ; WGP

Abstract

High dimensional genomic techniques have been used widely in recent years for association studies and for prediction of complex traits and diseases. Specific applications include genome-wide association studies (GWAS) for complex human diseases and genomic selection (GS) in animal and plant breeding. Prediction of complex traits and phenotypes using genetic data is an important goal, and while significant advances have been made, room for improvement remains before goals such as personalized medicine can be achieved. At the crux of the issue are statistical methods for creating prediction models from high-dimensional data. Whole-genome prediction models (WGP) have been highly successful in animal breeding, and some initial work has been done in the context of human disease. Techniques using variable selection from GWAS studies have been explored as well. We will discuss genetic and statistical factors affecting prediction accuracy (sample size, extent of linkage disequilibrium, genetic architecture and trait heritability), address ways of confronting complexity (gene x gene and gene x environmental interaction) and discuss advantages/disadvantages of model selection versus WGP.