A combination test for detection of gene-environment interaction in cohort studies

Brandon Coombes, Saonli Basu, Matt McGue

Research output: Contribution to journalArticlepeer-review

5 Scopus citations


Identifying gene-environment (G-E) interactions can contribute to a better understanding of disease etiology, which may help researchers develop disease prevention strategies and interventions. One big criticism of studying G-E interaction is the lack of power due to sample size. Studies often restrict the interaction search to the top few hundred hits from a genome-wide association study or focus on potential candidate genes. In this paper, we test interactions between a candidate gene and an environmental factor to improve power by analyzing multiple variants within a gene. We extend recently developed score statistic based genetic association testing approaches to the G-E interaction testing problem. We also propose tests for interaction using gene-based summary measures that pool variants together. Although it has recently been shown that these summary measures can be biased and may lead to inflated type I error, we show that under several realistic scenarios, we can still provide valid tests of interaction. These tests use significantly less degrees of freedom and thus can have much higher power to detect interaction. Additionally, we demonstrate that the iSeq-aSum-min test, which combines a gene-based summary measure test, iSeq-aSum-G, and an interaction-based summary measure test, iSeq-aSum-I, provides a powerful alternative to test G-E interaction. We demonstrate the performance of these approaches using simulation studies and illustrate their performance to study interaction between the SNPs in several candidate genes and family climate environment on alcohol consumption using the Minnesota Center for Twin and Family Research dataset.

Original languageEnglish (US)
Pages (from-to)396-412
Number of pages17
JournalGenetic epidemiology
Issue number5
StatePublished - Jul 2017


  • dimension reduction
  • gene-environment interaction
  • model selection
  • score tests

ASJC Scopus subject areas

  • Epidemiology
  • Genetics(clinical)


Dive into the research topics of 'A combination test for detection of gene-environment interaction in cohort studies'. Together they form a unique fingerprint.

Cite this