Using expression data in genetic association studies: An integrated bayesian approach to determine genetic pathways

Following the paradigm that the biological road from a genetic locus to the disease phenotype of interest leads through expression data, the availability of genetic data, expression profiles and the disease phenotype should allow for the identification of genetic pathways. However, statistical analysis approaches that integrate all three of these types of data are not straightforward. Given an association between the marker locus and the expression profile, and an association between the marker locus and the disease phenotype, we want to conclude that the genetic association with the phenotype is solely attributable to the genetic association with the expression profile in order to establish the pathway from gene to disease. As we show in this manuscript, this question can not be addressed using standard statistical methodology for hypothesis testing. We propose a Bayesian approach that can assess the genetic association with the phenotype, in the presence of the genomic association. Using simulation studies, we verify that the our approach has the desired properties. The approach is illustrated by an application to a real data set.