Mapping multiple complex and omics trait-associations using summary statistics from correlated or independent samples

发布时间 :2018-06-11  阅读次数 :623




主讲人:Dr. Lin Chen (陈琳)

Associate Professor, Department of Public Health Sciences, The University of Chicago

主讲人简介:陈琳博士2002年本科毕业于北京大学经济学院;2008年美国华盛顿大学获生物统计学博士学位;随后在普林斯顿大学、Fred Hutchinson癌症研究中心进行博士后研究工作;2010年起在芝加哥大学任助理教授、副教授。主要研究方向为统计基因组学与组学大数据统计分析,在Nature commutations、Genome Research、Genome Biology、Bioinformatics、 The American Journal of Human Genetics、Journal of American the Statistical Association等生物信息学与统计学领域重要期刊发表学术论文近30篇。


To date, genome-wide association studies (GWAS) have identified more than 60 thousand unique SNP-trait associations, but the functional mechanisms underlying many of these associations remain unknown. The analysis of the joint associations of a SNP to complex trait(s) and omics-phenotypes (e.g., mRNA expression, DNA methylation, and protein abundance) has the potential to elucidate mechanisms underlying known associations or to reveal novel relationships between genetic variants and complex traits.  In this work, we propose an integrative genomics approach for mapping multi-trait associations using summary statistics from correlated, overlapping or independent samples, and develop an R package: “primo.”  Different than existing literature, primo can analyze the joint associations of more than three sets of summary statistics. Primo estimates the probability that a SNP is associated with any arbitrary combination of (e.g. at least 1, at least 2, some but not others, or all) traits and omics-phenotypes of interest.  Empirical false discovery rates are estimated and controlled through permutation of test statistics.


We first applied primo to identify multi-omics QTLs in tumor samples from TCGA in breast tissues, and ovary tissues, separately. We joint analyzed the eQTL, meQTL, and pQTL summary statistics, and have identified multiple genes with e-me-pQTLs, several of which were known to be cancer-related.  We further integrated the three breast tissue QTL sets with summary statistics from a breast cancer GWAS, and identified several breast-cancer-risk-associated SNPs also being multi-omics QTLs. Those results illustrated that some SNPs, in particular those in non-coding regions, may affect cancer risks and other complex traits via certain omics phenotypes.  In another integrative analysis by primo, we estimated broad sharing (~73%) of eQTLs between tumor breast tissues from TCGA and normal breast tissues from GTEx.  Similar sharing (~78%) was estimated in the analysis of ovary tissues.  Finally, we used primo to assess cross-tissue mediation (trans-association) effects in multiple brain tissues from GTEx.   The package is freely available for download at

上海交通大学生命科学技术学院 Copyright © 2017 沪交ICP备05029. All Rights Reserved.