Predicting Complex Human Traits Using Whole-Genome Markers: Proof of Principle with Human Height and Human Longevity

• Activity Number: 566
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Gustavo de los Campos*+ and Yann Klimentidis and Robert Makowsky and Ana Ines Vazquez and Nicholas M. Pajewski and Christine Woods Duarte and David B. Allison
• Companies: University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham and Wake Forest University and University of Alabama at Birmingham and University of Alabama at Birmingham
• Keywords: Prediction ; SNPs ; Complex Traits

Abstract

Most diseases have a genetic component. However, despite great progress in genotyping technologies, our ability to predict genetic risk remains limited. A perhaps overlooked problem resides in limitations of the statistical methods used in genome-wide association studies. These methods, based on single marker associations, are most powerful for traits affected by a few genes, but are not well suited to complex traits. We applied whole-genome prediction (WGP), a predictive approach which uses dense single nucleotide polymorphism (SNPs) data, to predict human height (H) and years of life (YL). Data was from the Framingham Heart Study (N=5,117). Models with different marker density (from 2.5K to 400K SNPs) were fitted using different WGP methods. Using 400K SNPs we achieved a cross-validation-R2 (CV-R2) of 0.25 for (age and sex-adjusted) H. Previous studies with H have failed to achieve an R2>0.10 for H. For YL we achieved a CV-R2 of 0.21. The contribution of SNPs to CV-R2 in predicting YL was similar to that of sex, smoking and body-mass index combined. We conclude that WGP can offer opportunities to advance our ability to predict complex human traits.