fastASSET: Fast ASSET using pre-screening

fastASSET is an R package for joint genetic association analysis across a large number of traits. It is an accelerated version of the association analysis based on subsets (ASSET) method (Bhattacharjee et al, AJHG 2012). The input is GWAS summary statistics for one genetic variant across multiple traits, and the output is a p-value for global association and a list of associated traits. fastASSET accelerates the computation by restricting subset search among the traits with suggestive level of associations, defined by a liberal p-value threshold.

See the original ASSET R package for more information: https://bioconductor.org/packages/release/bioc/html/ASSET.html

Installation

First install devtools and ASSET

install.packages("devtools")
devtools::install_github("sbstatgen/ASSET")

then install fastASSET

devtools::install_github("gqi/fastASSET")

Type ?fastASSET::fast_asset() to view the documentation.

Step-by-step tutorial

1. Load example dataset.

library(fastASSET)
data("example_rs6678982", package="fastASSET")

Example dataset example_rs6678982 has been provided as part of the fastASSET package. It includes data of a single SNP required to run fastASSET :

SNP: SNP name
traits: Vector of trait names.
betahat: Effect of SNP on traits, obtained from GWAS summary statistics.
SE: Standard error of betahat.
Neff: Vector of effective sample size of GWAS. For continuous traits, the effective sample size is the total sample size; for binary traits, the effective sample size is Ncase*Ncontrol/(Ncase+Ncontrol).
ldscintmat: Matrix of bivariate LD score regression intercepts. It estimates the correlation of z-statistics across traits under the global null hypothesis.

We have provided ldscintmat with this example. In real data analysis, ldscintmat can be obtained by running bivariate LD score regression for each pair of traits. See the LDSC tutorial for running bivaraite LD score regression.

2. Create correlated trait blocks using hierarchical clustering

This step partitions the traits into smaller blocks correlated traits. Traits from independent blocks are treated as independent in subsequent analysis.

block <- create_blocks(ldscintmat)

3. Conduct fastASSET analysis

test <- fast_asset(snp=SNP, traits.lab=traits, beta.hat=betahat, sigma.hat=SE,
Neff=Neff, cor=ldscintmat, block=block, scr_pthr=0.05)

Reference

If you use this package or other custom scripts from this repository, please cite:

Qi, G., Chhetri, S. B., Ray, D., Dutta, D., Battle, A., Bhattacharjee, S.*, & Chatterjee, N.* (2022). Genome-Wide Large-Scale Multi-Trait Analysis Characterizes Global Patterns of Pleiotropy and Unique Trait-Specific Variants. bioRxiv. In press for Nature Communications.

*Corresponding authors.

See folders simulations and analysis for other custom scripts used in the paper.