光华讲坛——社会名流与企业家论坛第5653期

主题：Estimating Causal Relationship for Complex Traits with Weak and Heterogeneous Genetic Effects

主讲人：芝加哥大学王静姝博士

主持人：统计学院统计研究中心 林华珍教授

时间：2019年12月12日(星期三)下午4:00-5:00

地点：西南财经大学柳林校区弘远楼408会议室

主办单位：统计研究中心 统计学院 科研处

主讲人简介：

    Jingshu Wang joined The University of Chicago statistics department as an assistant professor in 2019. She received my Ph.D. in statistics from Stanford in 2016 and was a postdoc researcher at Wharton Statistics Department from 2016-2019. Her main research interest is in developing statistical methods for cutting-edge bio-technologies and genetic problems.

王静姝于2019年加入芝加哥大学统计学系，担任助理教授。她于2016年获得斯坦福大学统计学博士学位，并于2016-2019年担任沃顿商学院统计系博士后研究员。她的主要研究兴趣是发展尖端生物技术和遗传问题的统计方法。

主要内容：

Genetic association signals tend to be spread across the whole genome for complex traits. The recently proposed ''omnigenic'' model indicates that, when the risk factor is a complex trait, most genetic variants can weakly affect the risk factor while also easily affecting a common disease not through the risk factor. Existing methods in Mendelian Randomization (MR) are not ideal under such pervasive pleiotropy. We propose a comprehensive framework GRAPPLE (Genome-wide mR Analysis under Pervasive PLEiotropy) for MR, utilizing both strongly and weakly associated genetic variant, and can detect the existence of multiple pleiotropic pathways. We show that GRAPPLE is a comprehensive tool that can detect and adjust for pleiotropy, simultaneously estimate the causal effect of multiple risk factors and can determine the causal relationship direction using three-sample GWAS summary statistics datasets. With GRAPPLE, we conduct a screening for the lipid traits (HDL-C, LDL-C and triglycerides) with around 30 common diseases to understand their roles as risk factors and detect potential pleiotropic pathways.

对于复杂的性状，遗传关联信号往往会在整个基因组中传播。最近提出的“全基因组”模型表明，当风险因子是一种复杂的性状时，大多数遗传变异对风险因子的影响较弱，而对普通疾病的影响很容易不通过危险因子。现有的孟德尔随机化(MR)方法在这种普遍的多效性下并不理想。我们提出了一个适用于MR的综合框架GRAPPLE（普适基因多效型下的全基因组mR分析），它利用了强弱关联的遗传变异，并且可以检测到多种多效性途径的存在。结果表明，GRAPPLE是一个综合工具，可以检测和调整多效性，同时估计多个风险因子的因果关系，并可以使用三样本GWAS汇总统计数据集确定因果关系方向。通过GRAPPLE，我们对大约30种常见疾病的脂质特征（HDL-C，LDL-C和甘油三酸酯）进行了筛选，以了解其作为风险因子的作用并检测潜在的多效性途径。