Abstract:
|
There is increasing interest in detecting gene-¬-by-¬-gene interactions for complex traits, with varying, but substantial, proportions of heritability remaining unexplained by surveys of single-¬-SNP genetic variation. The major challenges from traditional regression-¬-based methods are the large number of possible pairs under investigation, with a requisite need to correct for multiple testing, and the restrictive assumptions of large marginal effects to reduce the search space and limit the number of tests. Both of these challenges may limit power, especially when the marginal effects are in fact modest. In this talk, we propose a new procedure for detecting gene-¬-by-¬-gene interactions through meta-¬-analyses. Our approach is pragmatic when data-¬-sharing limitations restrict mega-¬-analyses. It is also computationally efficient in that it applies a dimension reduction procedure and thus may scale for higher-¬-order interactions as well. We compare the type I error and power of our proposed procedure relative to existing methods and evaluate their strengths and limitations.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.