A1093
Title: Quantifying the global mediation effect for nonsparse high dimensional genomics mediators
Authors: Chunlin Li - Iowa State University (United States) [presenting]
Tianzhong Yang - University of Minnesota (United States)
Abstract: While many existing epidemiological studies have examined associations between alcohol and cardiovascular outcomes, less has been done to explore causal biological pathways and mechanisms of the observed associations at the molecular level. To investigate this relationship, we propose a new causal measure to quantify the mediating role of molecular phenotypes, such as DNA methylation, in bridging alcohol intake and cardiovascular outcomes. The challenge of estimating this measure is two-fold. First, since alcohol consumption is associated with genome-wide changes at the molecular level, it is biologically plausible that many omics mediators with weak but collectively considerable effects are involved in the pathway; however, existing methods are plagued by inconsistency in the presence of non-sparse mediators. To address this issue, we develop a method to estimate the proposed measure in such situations consistently. Second, many epidemiological studies use case-control sampling, which introduces ascertainment bias in mediation analysis. To correct this bias, we propose a method of moment motivated by heritability estimation. Finally, a significant challenge in this research is the potential for residual confounding in observational studies, which can seriously compromise the validity of scientific findings. We will briefly discuss the approach to correct the confounding bias.
