Assessing the contribution of rare variants to complex trait heritability from whole-genome sequence data

Abstract

Analyses of data from genome-wide association studies on unrelated individuals have shown that, for human traits and diseases, approximately one-third to two-thirds of heritability is captured by common SNPs. However, it is not known whether the remaining heritability is due to the imperfect tagging of causal variants by common SNPs, in particular whether the causal variants are rare, or whether it is overestimated due to bias in inference from pedigree data. Here we estimated heritability for height and body mass index (BMI) from whole-genome sequence data on 25,465 unrelated individuals of European ancestry. The estimated heritability was 0.68 (standard error 0.10) for height and 0.30 (standard error 0.10) for body mass index. Low minor allele frequency variants in low linkage disequilibrium (LD) with neighboring variants were enriched for heritability, to a greater extent for protein-altering variants, consistent with negative selection. Our results imply that rare variants, in particular those in regions of low linkage disequilibrium, are a major source of the still missing heritability of complex traits and disease.

Figure 1:

GREML-LDMS estimates with 8 bins (2 LD bins for each of the 4 MAF bins) correcting for 20 PCs (calculated from LD-pruned HM3 SNPs) after imputing SNPs from Illumina InfiniumCore24, GSA 24 and Affymetrix Axiom arrays using Haplotype Reference Consortium reference panels for N=25,465 samples. (A) Estimates of $h_{G+\mathit{IMP}}^2$ for height are between 0.50-0.56 (SE 0.06-0.07). (B) Estimates for BMI are between 0.16-0.21 (SE 0.07). The large SEs of the estimates for variants with MAF between 0.0001 to 0.001 can be explained by the large number of imputed variants in this MAF bin because the sampling variance of a SNP-based heritability estimate is proportional to the effective number of independent variants . Between ~19.0M and ~20.0M variants in total are included in the analysis. The number of variants in each of the 4 MAF bins (twice the number in each LD bin) can be found in Supplementary Figure 8.

Figure 2:

GREML-LDMS of height and BMI for N=25,465 samples using 3 or 4 LD groups for each MAF bin correcting for 48/160/320 PCs computed from WGS variants. Each variant was allocated in a tertile or a quartile according to its LD score. (A) Estimates using 3 LD bins for height: 0.68 (SE 0.09 – 0.10). (B) Estimates using 3 LD bins for BMI: 0.30 – 0.32 (SE 0.10). (C) Estimates using 4 LD bins for height: 0.67 – 0.68 (SE 0.10). (D) Estimates using 4 LD bins for BMI: 0.28 – 0.30 (SE 0.10).

Figure 3:

Variance explained per variant (the estimate of genetic variance divided by the number of variants in each bin) from GREML-LDMS with the low-LD and low-MAF (< 0.1) variants partitioned into 2 distinct categories according to the SnpEff putative effect of the variant (protein-altering or non-protein-altering), correcting for 48 PCs from WGS variants for N=25,465 samples. There is a total of 11 genetic components in this analysis. There is an apparent enrichment of heritability in the protein-altering groupings (low LD) over non-protein-altering (low LD) or high LD variants for height (A) as well as for BMI, although the standard errors for this trait are large (B).