Genome-wide compound heterozygote analysis highlights alleles associated with adult height in Europeans

Abstract

Adult height is the most widely genetically studied common trait in humans; however, the trait variance explainable by currently known height-associated single nucleotide polymorphisms (SNPs) identified from the previous genome-wide association studies (GWAS) is yet far from complete given the high heritability of this complex trait. To exam if compound heterozygotes (CH) may explain extra height variance, we conducted a genome-wide analysis to screen for CH in association with adult height in 10,631 Dutch Europeans enriched with extremely tall people, using our recently developed method implemented in the software package CollapsABEL. The analysis identified six regions (3q23, 5q35.1, 6p21.31, 6p21.33, 7q21.2, and 9p24.3), where multiple pairs of SNPs as CH showed genome-wide significant association with height (P < 1.67 × 10^-10). Of those, 9p24.3 represents a novel region influencing adult height, whereas the others have been highlighted in the previous GWAS on height based on analysis of individual SNPs. A replication analysis in 4080 Australians of European ancestry confirmed the significant CH-like association at 9p24.3 (P < 0.05). Together, the collapsed genotypes at these six loci explained 2.51% of the height variance (after adjusting for sex and age), compared with 3.23% explained by the 14 top-associated SNPs at 14 loci identified by traditional GWAS in the same data set (P < 5 × 10^-8). Overall, our study empirically demonstrates that CH plays an important role in adult height and may explain a proportion of its "missing heritability". Moreover, our findings raise promising expectations for other highly polygenic complex traits to explain missing heritability identifiable through CH-like associations.

Fig. 1

Manhattan plots from RS. Minus log of P value is represented by blue and red dots and grouped by chromosome. a Manhattan plot for single-SNP GWAS. The horizontal line (−log 5 × 10⁻⁸) indicates threshold of genome-wide significance. b Manhattan plot for GCDH scan, the horizontal line (−log 1.67 × 10⁻¹⁰) indicates threshold of GCDH genome-wide significance. Loci are annotated in green if they have not been reported before with genome-wide height association by the GIANT consortium (Wood et al. 2014)

Fig. 2

QQ plots. a QQ plot of single-SNP P values. b QQ plot of GCDH P values. In both cases, P values from five permutation analyses were also plotted to estimate the null distribution

Fig. 3

Regional Manhattan plot of GCDH P values from 9p24.3