Despite current enthusiasm for investigation of gene-gene and gene-environment interactions, very few discoveries have been published. The identification of interactions or at least genetic and non-genetic factors involved in these interactions remains one of the greatest challenges in genetic epidemiology. Multiple testing issues, unavailability of environmental exposure and high-order interactions are commonly cited as fundamental problems in identifying these interactions. However alternative strategies are possible in the case of quantitative phenotypes when the aim is not to identify interaction effect per se but rather to detect quantitative trait loci (QTLs) involved in interaction. Most association tests of marginal effect are designed to capture a shift in phenotypic values on a particular location (e.g. the phenotypic mean) in individuals carrying a risk allele as compared to individuals that do not. However, in a situation where a QTL interacts with unknown risk factors, phenotypic differences between carrier and non-carrier can be expressed by differences in the shape of the distribution of phenotypic values rather than differences on a single location. We propose a non-parametric test of association that leverages such information. By construction, the test uses only the phenotypic and genotypic information and does not require information on the (possibly unknown) interacting factors. A preliminary study on simulated data shows that the proposed test can be more powerful than the test of the standard test of marginal linear effect in the presence of a strong interaction effect or in the presence of stratum-specific effects in opposite directions. In this project, we aim first to condut additional theoretical work including power comparison with other methods under different interaction models and identification of technical improvements. Second, the proposed approach will be applied in real genome-wide data to identify genetic variants associated with four biomarkers in the one-carbon metabolism pathway. Finally, extension of this method for multi-markers analysis will be explored. This method development project is responding to the need for new and innovative methods for detection of genetic variants involved in interactions that has been recently highlighted by multiple reports including a recent NIH workshop. The characterization and the application of the proposed method in GWAS data will be a step toward the discovery of these variants.

Public Health Relevance

Epistasis and gene-environment interactions are two potential keys that may explain an important part of the so called 'missing heritability'of complex traits. Despite numerous studies, very few discoveries have been published and the identification of genetic and non-genetic risk factors involved in these interactions remains one of the greatest challenges in genetic epidemiology. In this research proposal, we propose to develop and apply a new and very innovative approach that aims to detect genetic variants involved in interactions.

National Institute of Health (NIH)
National Human Genome Research Institute (NHGRI)
Small Research Grants (R03)
Project #
Application #
Study Section
Genomics, Computational Biology and Technology Study Section (GCAT)
Program Officer
Ramos, Erin
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Harvard University
Public Health & Prev Medicine
Schools of Public Health
United States
Zip Code
Aschard, Hugues; Vilhjálmsson, Bjarni J; Joshi, Amit D et al. (2015) Adjusting for heritable covariates can bias effect estimates in genome-wide association studies. Am J Hum Genet 96:329-39
Aschard, Hugues; Gusev, Alexander; Brown, Robert et al. (2015) Leveraging local ancestry to detect gene-gene interactions in genome-wide data. BMC Genet 16:124
Aschard, Hugues; Vilhjálmsson, Bjarni J; Greliche, Nicolas et al. (2014) Maximizing the power of principal-component analysis of correlated phenotypes in genome-wide association studies. Am J Hum Genet 94:662-76