Doctoral candidate in Biostatistics, Yangqing Deng, will present:
“Genetic Testing with Conditional Analysis and Summary Statistics”
PhD Adviser: Wei Pan
Abstract: There is a growing interest in testing genetic pleiotropy, meaning that a genetic variant is associated with multiple traits, which may help us know more about the biological mechanisms and can lead to some medical applications including drug repurposing. Many methods proposed to test pleiotropy use individual level data and marginal analysis, while in many situations we are only able to obtain summary statistics instead of individual level data, and marginal analysis cannot distinguish direct and indirect effects. It is useful to consider conditional analysis with summary statistics, where a subset of traits are adjusted for another subset of traits. For our first project, we propose our method to build conditional models using summary statistics only, and we use simulations and real data applications to demonstrate the advantage of our method as well as how marginal and conditional analyses are different. For our second project, we develop a new pleiotropy testing procedure based on summary statistics that can be applied for both marginal analysis and conditional analysis, based on the idea of union-intersection testing as in Schaid et al. (2016). We also propose another version of the test using generalized estimating equations (GEE) under the working independence model for robust inference. We provide numerical examples to show the effectiveness of our new methods. For our third project, we move from testing pleiotropy to testing colocalization, where the major difference is that pleiotropy is defined as a genetic variant is significantly associated with more than one of the traits of interest, while colocalization is defined as at least one variant in a locus is significantly associated with all of the traits of interest. We propose our new methods to address the issues of existing methods, and use simulations and real data examples to confirm the advantages of our methods.