Sorry for the naivety, but what can we use to capture the effect of rare but low-effect size variants, too rare for GWAS but without enough signal to show up in exome sequencing studies? Alternatively, can we estimate indirectly their contribution to missing h^2?
Thanks!
Good question. For detection of individual effect sizes of small effect & rare variants, our only real strategy (AFAIK) is to increase sample size. For the most part, the rarest of such variants will remain hidden from us due to lack of power.
With respect to whether we can estimate their overall contribution to heritability, we can use variants of GREML (genomic REML) to estimate their overall contribution to SNP-heritability. E.g., we can use GRMs binned by MAF and LD from sequence or imputed data and fit all the multiple GRMs simultaneously. In theory, this should capture increasingly rare heritability (e.g., see Evans et al., 2018 - Comparison of methods that use whole genome data to estimate the heritability and genetic architecture of complex traits | Nature Genetics). In practice (e.g., Wainschtein et al, 2022 - Assessing the contribution of rare variants to complex trait heritability from whole-genome sequence data | Nature Genetics), it is not completely clear that we have adequately controlled for population stratification for rare variant heritability, which is where Ben’s skepticism comes from
2 Likes
Thanks, Giulio for the question. Burden tests (or any other way to aggregate rare variants) are exactly trying to address this question. Wei was showing her analysis of BRCA1/2 in the UKB, where you can only pick that signal by aggregating SNPs within the genes, while single variant analyses are clearly unpowered.
1 Like