Adjustment for covariates using summary statistics of genome-wide association studies

Tao Wang, Xiaonan (Nan) Xue, Xianhong Xie, Qian K. Ye, Xiaofeng Zhu, Robert C. Elston

Research output: Contribution to journalArticle

Abstract

Linear regression is a standard approach to identify genetic variants associated with continuous traits in genome-wide association studies (GWAS). In a standard epidemiology study, linear regression is often performed with adjustment for covariates to estimate the independent effect of a predictor variable or to improve statistical power by reducing residual variability. However, it is problematic to adjust for heritable covariates in genetic association analysis. Here, we propose a new method that utilizes summary statistics of the covariate from additional samples for reducing the residual variability and hence improves statistical power. Our simulation study showed that the proposed methodology can maintain a good control of Type I error and can achieve much higher power than a simple linear regression. The method is illustrated by an application to the GWAS results from the Genetic Investigation of Anthropometric Traits consortium.

Original languageEnglish (US)
JournalGenetic Epidemiology
DOIs
StateAccepted/In press - Jan 1 2018

Fingerprint

Genome-Wide Association Study
Linear Models
Epidemiology

Keywords

  • continuous traits
  • genome-wide association studies (GWAS)
  • linear regression, pleiotropy
  • meta-analysis

ASJC Scopus subject areas

  • Epidemiology
  • Genetics(clinical)

Cite this

@article{cbee312a5bff47b1b2558c451cd595bf,
title = "Adjustment for covariates using summary statistics of genome-wide association studies",
abstract = "Linear regression is a standard approach to identify genetic variants associated with continuous traits in genome-wide association studies (GWAS). In a standard epidemiology study, linear regression is often performed with adjustment for covariates to estimate the independent effect of a predictor variable or to improve statistical power by reducing residual variability. However, it is problematic to adjust for heritable covariates in genetic association analysis. Here, we propose a new method that utilizes summary statistics of the covariate from additional samples for reducing the residual variability and hence improves statistical power. Our simulation study showed that the proposed methodology can maintain a good control of Type I error and can achieve much higher power than a simple linear regression. The method is illustrated by an application to the GWAS results from the Genetic Investigation of Anthropometric Traits consortium.",
keywords = "continuous traits, genome-wide association studies (GWAS), linear regression, pleiotropy, meta-analysis",
author = "Tao Wang and Xue, {Xiaonan (Nan)} and Xianhong Xie and Ye, {Qian K.} and Xiaofeng Zhu and Elston, {Robert C.}",
year = "2018",
month = "1",
day = "1",
doi = "10.1002/gepi.22148",
language = "English (US)",
journal = "Genetic Epidemiology",
issn = "0741-0395",
publisher = "Wiley-Liss Inc.",

}

TY - JOUR

T1 - Adjustment for covariates using summary statistics of genome-wide association studies

AU - Wang, Tao

AU - Xue, Xiaonan (Nan)

AU - Xie, Xianhong

AU - Ye, Qian K.

AU - Zhu, Xiaofeng

AU - Elston, Robert C.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Linear regression is a standard approach to identify genetic variants associated with continuous traits in genome-wide association studies (GWAS). In a standard epidemiology study, linear regression is often performed with adjustment for covariates to estimate the independent effect of a predictor variable or to improve statistical power by reducing residual variability. However, it is problematic to adjust for heritable covariates in genetic association analysis. Here, we propose a new method that utilizes summary statistics of the covariate from additional samples for reducing the residual variability and hence improves statistical power. Our simulation study showed that the proposed methodology can maintain a good control of Type I error and can achieve much higher power than a simple linear regression. The method is illustrated by an application to the GWAS results from the Genetic Investigation of Anthropometric Traits consortium.

AB - Linear regression is a standard approach to identify genetic variants associated with continuous traits in genome-wide association studies (GWAS). In a standard epidemiology study, linear regression is often performed with adjustment for covariates to estimate the independent effect of a predictor variable or to improve statistical power by reducing residual variability. However, it is problematic to adjust for heritable covariates in genetic association analysis. Here, we propose a new method that utilizes summary statistics of the covariate from additional samples for reducing the residual variability and hence improves statistical power. Our simulation study showed that the proposed methodology can maintain a good control of Type I error and can achieve much higher power than a simple linear regression. The method is illustrated by an application to the GWAS results from the Genetic Investigation of Anthropometric Traits consortium.

KW - continuous traits

KW - genome-wide association studies (GWAS)

KW - linear regression, pleiotropy

KW - meta-analysis

UR - http://www.scopus.com/inward/record.url?scp=85053492335&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053492335&partnerID=8YFLogxK

U2 - 10.1002/gepi.22148

DO - 10.1002/gepi.22148

M3 - Article

JO - Genetic Epidemiology

JF - Genetic Epidemiology

SN - 0741-0395

ER -