Testing Multiple Biological Mediators Simultaneously

• Activity Number: 624
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Simina Boca*+ and Joshua Sampson
• Companies: National Cancer Institute and National Cancer Institute
• Keywords: multiple comparisons ; permutation test ; genomics ; metabolomics

Abstract

Understanding the biological pathways between exposures and outcomes is crucial in many epidemiology applications. An exposure can have direct and mediated effects on a disease, with one or more mediators present. For example, one may be interested in whether a particular genotype affects a phenotype as a result of changes in the expression of one or more genes. There are many methods that can test whether a putative mediator lies in a pathway. However, there is little discussion on how to handle the case when there are multiple putative mediators, as may be the case in the area of metabolomics. We develop a permutation-based method for testing for the presence of multiple pathway mediators, controlling the family-wise error rate. The null hypothesis is that there is no indirect effect through a mediator: There may be no causal effect from the mediator to the disease, from the exposure to the mediator, or from both the mediator to the disease and the exposure to the mediator. We evaluate our method through simulations and on genomics and metabolomics datasets.