Regional heritability mapping method helps explain missing heritability of blood lipid traits in isolated populations.

M Shirali,C S Haley,C Hayward,I Rudan,R Pong-Wong,P Navarro,H Campbell,S Knott,V Vitart,A F Wright,N D Hastie

doi:10.1038/hdy.2015.107

Abstract

Single single-nucleotide polymorphism (SNP) genome-wide association studies (SSGWAS) may fail to identify loci with modest effects on a trait. The recently developed regional heritability mapping (RHM) method can potentially identify such loci. In this study, RHM was compared with the SSGWAS for blood lipid traits (high-density lipoprotein (HDL), low-density lipoprotein (LDL), plasma concentrations of total cholesterol (TC) and triglycerides (TG)). Data comprised 2246 adults from isolated populations genotyped using ∼300 000 SNP arrays. The results were compared with large meta-analyses of these traits for validation. Using RHM, two significant regions affecting HDL on chromosomes 15 and 16 and one affecting LDL on chromosome 19 were identified. These regions covered the most significant SNPs associated with HDL and LDL from the meta-analysis. The chromosome 19 region was identified in our data despite the fact that the most significant SNP in the meta-analysis (or any SNP tagging it) was not genotyped in our SNP array. The SSGWAS identified one SNP associated with HDL on chromosome 16 (the top meta-analysis SNP) and one on chromosome 10 (not reported by RHM or in the meta-analysis and hence possibly a false positive association). The results further confirm that RHM can have better power than SSGWAS in detecting causal regions including regions containing crucial ungenotyped variants. This study suggests that RHM can be a useful tool to explain some of the ‘missing heritability' of complex trait variation.

Highlights

Blood lipoprotein concentration is an important risk factor for coronary heart disease and stroke (Lewington et al, 2007)
regional heritability mapping (RHM) vs SSGWAS Table 2 presents the estimated variances of the significant single-nucleotide polymorphism (SNP) detected by the SSGWAS method and their regional variance estimated by the RHM method compared with the SNP variances suggested by meta-analysis
This SNP and the region containing it was not detected in the meta-analysis of Teslovich et al (2010) and in the most current meta-analysis report in high-density lipoprotein (HDL) by Surakka et al (2015) and no report was found for any significant SNP around this region in genome-wide association studies (GWASs) catalog for HDL

Summary

Introduction

Blood lipoprotein concentration is an important risk factor for coronary heart disease and stroke (Lewington et al, 2007). In most GWASs, single single-nucleotide polymorphism (SNP) analysis (SSGWAS) has been used to determine associated variants and subsequently their contribution to complex trait variation This method has been shown to lack power for the detection of rare genetic variants (Bodmer and Tomlinson, 2010) because of low linkage disequilibrium between rare alleles and genotyped SNPs on the SNP genotyping array (Zeggini et al, 2005). Uemoto et al (2013) demonstrated the advantage in power of RHM in comparison with SSGWAS and some gene-based association methods, such as versatile gene-based association study (Liu et al, 2010), SNP-set (Sequence) Kernel Association Test (Wu et al, 2011) and canonical correlation analysis (Tang and Ferreira, 2012), using simulations based on real genotype data from a human population and a wide range of scenarios for quantitative trait loci (QTLs) with both common and rare alleles. SNP-set (Sequence) Kernel Association Test power can be improved by changing the β-weights (Belonogova et al, 2013), and Received 13 May 2015; revised 23 October 2015; accepted 26 October 2015; published online 23 December 2015 specific family-based versions of the approach are available (Chen et al, 2013; Svishcheva et al, 2014)

Objectives

Methods

Results

Discussion

Conclusion