Skip to content

Home

Jump to bottom
Kathy Burch edited this page Sep 5, 2019 · 9 revisions

Welcome to the h2-GRE wiki! In the following sections, we describe how to use h2-GRE to estimate total (genome-wide) SNP-heritability.

In order for the desirable properties of the h2-GRE estimator to hold, h2-GRE requires individual-level genotype and phenotype data where the number of individuals (N) is larger than the number of genotyped SNPs per chromosome (m_k for the k-th chromosome). As a reference point, in our analyses of the UK Biobank genotyped SNPs (UK Biobank Axiom Array), chromosome 1 had about 45-55K SNPs depending on how we QC-ed the data.

Estimating h2-GRE consists of 3 main steps:

  1. Compute chromosome-wide LD.
  2. Compute the inverse (or pseudoinverse) of each chromosome-wide LD matrix.
  3. Perform GWAS (i.e. OLS regression) and compute h2-GRE.

The second step (computing inverse LD) is the computational bottleneck; the other steps can be parallelized. Details on each step can be found in the following sections and in Hou*, Burch*, et al. Nat Genet, 2019.

Clone this wiki locally