Developing Polygenic Risk Prediction Model Based on Winner's Curse Correction and Multidimensional Thresholding

• Activity Number: 153
• Type: Invited
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: WNAR
• Author(s): Jianxin Shi* and Nilanjan Chatterjee
• Companies: National Cancer Institute and National Cancer Institute
• Keywords: Heirtability ; Lasso ; eQTL ; mQTL

Abstract

Estimates of heritability suggest that genome-wide association studies have the potential to improve genetic risk prediction models, but SNPs discovered at genome-wide significance level typically have low discriminatory power. We will explore methods for optimal thresholding and weighting SNPs to improve performance of models exploiting heritability in SNPs that do not achieve genomewide significance levels. In particular, we will study effects of winner's curse correction in estimation of weights for the SNPs and use of functional/annotation information for defining variable thresholds for SNP selection. Methods will be illustrated based on models built from summary-level results from large GWAS for a variety of complex traits.