While genetic consortia have identified thousands of genetic variants influencing hundreds of complex human diseases, the underlying mechanisms by which such variants influence these disorders remain largely unexplored. There is substantial interest in exploring how the relationship between genetic variation and outcomes of interest are influenced by intermediate biological, environmental, and phenotypic factors. Our own group is involved in a large GWAS study of orofacial clefting that seeks to understand how variants increase susceptibility to cleft lip/palate and whether intermediate risk factors like tobacco/alcohol use play a mechanistic role. In general, we can explore these relationships using mediation techniques, which are invaluable for disentangling relationships between exposures and outcomes and assessing mechanisms by which such relationships are influenced by intermediate mediators. Our orofacial GWAS study, like most genetic studies, employs a case-control study design. Under this design, the most common procedure for mediation analysis is based on a counterfactual framework developed by VanderWeele and colleagues. This framework showed how indirect and direct effects for dichotomous outcomes could be estimated within case-control studies on the odds ratio scale, allowing for exposure- mediator interactions and nonlinear effects. While this framework has proven valuable within the scientific community (>1100 citations), we show it does not fully leverage all relevant data collected by the case-control study. This can lead to the framework having sub-optimal performance; possibly leading to an increase in type II errors for testing significant relationships among exposures, mediators, and outcomes in case-control datasets like the orofacial GWAS that is the focus of our proposal. In this proposal, we will develop novel likelihood-based approaches for mediation analysis in case-control studies that fully leverages all available data, thereby leading to more precise estimates and more powerful testing of indirect and direct effects compared to existing frameworks. Once developed, we will apply these methods to the orofacial clefting GWAS to refine relationships between risk SNPs and cleft lip/palate and deduce whether these relationships are mediated by variables like maternal tobacco use during pregnancy. We further will create public software implementing these approaches, which will facilitate wide dissemination to the scientific community. Our resulting work and software will have substantial impact both in genetic studies of complex human traits as well as in broader topics across Public Health.
This project will develop improved tools for mediation analyses in case-control studies. This project will further apply these tools to a case-control genome-wide association study of orofacial clefting to elucidate the genetic mechanisms underlying this debilitating disorder. We will also create public software that will facilitate wide dissemination of these tools, thereby enabling their use by the scientific community at large.
