时间：2018年6月25日上午8:30-9:20
地点：法商南楼135室（闵行校区）
报告人：王静姝（宾夕法尼亚大学）
Title: A comprehensive Framework for Genome-wide Mendelian Randomization under Pervasive Pleiotropy
Abstract:
Mendelian randomization (MR) is a method of exploiting genetic variation to unbiasedly estimate a causal effect in presence of unmeasured confounding. MR is being widely used in epidemiology and other related areas of population science. One major challenge of applying MR to infer the causal relationship between complex traits is that the selected genetic variations can easily become invalid Instrumental Variables (IVs) with the presence of pleiotropy, which has recently been shown to be pervasive for complex traits.
Existing methods in MR can easily become invalid under such pervasive pleiotropy. We propose a comprehensive framework, GRAPPLE, for MR using GWAS data, which brings up the importance of using a large number of weakly effective genetic variants. With theoretical guarantees, RAPS can correctly estimate the causal relationship under With theoretical guarantees, GRAPPLE can correctly estimate the causal relationship under several common pleiotropy scenarios. Moreover, GRAPPLE can detect the existence of multiple pleiotropic pathways. We evaluate the performance of GRAPPLE using several case studies and show that it is a reliable method. As an application of GRAPPLE, we find that HDL-C has a protective effect on chronic artery disease after adjusting for other lipids including LDL-C and triglycerides if the weakly associated variants are utilized.
Short bio:
Jingshu Wang is a Postdoc-Researcher in statistics at the Department of Statistics, the Wharton School, University of Pennsylvania. She obtained her Ph.D. in statistics from Stanford University. Her research has been focused on high dimensional factor models, multiple hypotheses testing, meta-analysis, causal inference and statistical genetics. Her current work involves developing statistical methods for several cutting-edge biotechnologies and genetic problems, including single-cell RNA sequencing, genome-wide Mendelian Randomization, structural variation in the 3D genome, and high-throughput DNA-barcoded CRISPR screens in yeast for neurodegenerative diseases.