h2gene

h2gene is a method for partitioning gene-level contributions to complex-trait heritability by allele frequency (or any other binary annotations). It requires variant-level summary association statistics and in-sample LD (that is, LD estimated from a subset of the individuals in the association study).

Details about h2gene and our application of the method to 17,437 genes and 25 quantitative traits in the UK Biobank can be found in our paper:

Burch, Hou, et al. "Partitioning gene-level contributions to complex-trait heritability by allele frequency identifies disease-relevant genes." AJHG (2022).

h2gene estimates for 17,437 genes and 25 traits available for download

Estimates of h2gene and MAF-partitioned h2gene for the 17,437 genes and 25 traits analyzed in Burch, Hou, et al. (2022) are available for download here.

• H2GENE_ALL_MEAN is the posterior mean of total h2gene
• H2GENE_RARE_MEAN is the posterior mean of the rare variant component of h2gene (0.5% < MAF < 1%)
• H2GENE_LF_MEAN is the posterior mean of the low-frequency variant component of h2gene (1% < MAF < 5%)
• H2GENE_COMMON_MEAN is the posterior mean of the common variant component of h2gene (MAF > 5%)
• H2GENE_*_SD are the posterior standard deviations corresponding to each estimand (ALL, RARE, LF, and COMMON)
• H2GENE_*_LOWER and H2GENE_*_UPPER are the lower and upper bounds of the 90% credible intervals for each estimand (ALL, RARE, LF, and COMMON)
• *_CAUSAL_NUM_MEAN (and *_CAUSAL_NUM_SD) are the expected numbers of causal variants contributing to each h2gene estimate (and corresponding standard deviations)
• The full names of the UKB traits in TRAIT can be found here.

Metadata for the genes (chromosome, gene start/stop positions, number of variants in estimand, etc.) are also provided here.

Installation

# install.packages("remotes")
remotes::install_github("bogdanlab/h2gene")

Quick start

library(susieR)
library(h2geneR)
library(matrixStats)

# a minimal example from susieR website
set.seed(1)
n    <- 1000
p    <- 500
beta <- rep(0,p)
beta[c(1,2,300,400)] <- 1
X   <- matrix(rnorm(n*p),nrow=n,ncol=p)
y   <- X %*% beta + rnorm(n)
susie_fit <- susie(X,y,L=10)

# calculate LD matrix
ld <- cov(X)

# build annotation matrix
annot <- matrix(F, nrow=p, ncol=2)
colnames(annot) <- c("gene1", "gene2")
annot[c(1, 2), "gene1"] <- T
annot[c(300), "gene2"] <- T

# run h2gene
res <- h2gene(susie_fit, ld=ld, annot=annot)

# summarize h2gene results
print("Mean of heritability estimates across posterior samples:")
print(colMeans(res$hsq))

print("Standard deviations:")
print(colSds(res$hsq))

Paper experiments

See paper/ directory for code to replicate experiment results.