Activity Number:
|
252
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 1, 2016 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
Abstract #319707
|
View Presentation
|
Title:
|
Genomic Analysis with Common Binary Outcomes via Mediation
|
Author(s):
|
Sheila Gaynor* and Xihong Lin
|
Companies:
|
Harvard and Harvard T.H. Chan School of Public Health
|
Keywords:
|
genetics ;
genomics ;
mediation ;
integrative analysis
|
Abstract:
|
Mediation methods have been developed to decompose a total effect into natural direct and indirect effects in many settings. Modeling assumptions are made in order to utilize simple regression for identifying these effects; binary outcomes are regularly assumed to be rare or to be related through a mediator with a Bridge distribution. There are many settings where this simplifying assumption is not appropriate, particularly when analyzing genomic data for common tumor subtypes within a cancer type. Thus we propose a regression-based method that relaxes the rarity assumption for studies with a binary outcome and continuous mediator by exploiting the relationship between logit and probit models. We derive closed-form expressions for the natural direct and indirect effects on the odds ratio scale as well as statistics for testing. Further extensions are given for the case-control study setting. This method is assessed via high dimensional simulations. The method's performance at identifying effects of interest is demonstrated for cancer samples with multiple genomic datatypes from The Cancer Genome Atlas.
|
Authors who are presenting talks have a * after their name.