Powerful Set-Based Gene-Environment Interaction Testing Framework for Complex Diseases.

Authors

Jiao S, Peters U, Berndt S, Bézieau S, Brenner H, Campbell PT, Chan AT, Chang-Claude J, Lemire M, Newcomb PA, Potter JD, Slattery ML, Woods MO, Hsu L

Affiliations

• Public Health Sciences Division, Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America.
• Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, Maryland, United States of America.
• Service de Génétique Médicale, CHU Nantes, Nantes, France.
• Division of Clinical Epidemiology and Aging Research, German Cancer Research Center, Heidelberg, Germany.
• Epidemiology Research Program, American Cancer Society, Atlanta, Georgia, United States of America.
• Channing Division of Network Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts, United States of America.
• Division of Cancer Epidemiology, German Cancer Research Center, Heidelberg, Germany.
• Ontario Institute for Cancer Research, Toronto, Canada.
• Department of Internal Medicine, University of Utah Health Sciences Center, Salt Lake City, Utah, United States of America.
• Discipline of Genetics, Memorial University of Newfoundland, St. John's, NL, Canada.

Abstract

Identification of gene-environment interaction (G × E) is important in understanding the etiology of complex diseases. Based on our previously developed Set Based gene EnviRonment InterAction test (SBERIA), in this paper we propose a powerful framework for enhanced set-based G × E testing (eSBERIA). The major challenge of signal aggregation within a set is how to tell signals from noise. eSBERIA tackles this challenge by adaptively aggregating the interaction signals within a set weighted by the strength of the marginal and correlation screening signals. eSBERIA then combines the screening-informed aggregate test with a variance component test to account for the residual signals. Additionally, we develop a case-only extension for eSBERIA (coSBERIA) and an existing set-based method, which boosts the power not only by exploiting the G-E independence assumption but also by avoiding the need to specify main effects for a large number of variants in the set. Through extensive simulation, we show that coSBERIA and eSBERIA are considerably more powerful than existing methods within the case-only and the case-control method categories across a wide range of scenarios. We conduct a genome-wide G × E search by applying our methods to Illumina HumanExome Beadchip data of 10,446 colorectal cancer cases and 10,191 controls and identify two novel interactions between nonsteroidal anti-inflammatory drugs (NSAIDs) and MINK1 and PTCHD3.