光华讲坛——社会名流与企业家论坛第 5681 期
主题：An integrative pipeline of multi-tissue multi-trait Transcriptome-wide Association Study.
主办单位：统计研究中心 统计学院 科研处
冯贺莲于2017年获得哈佛大学公共卫生学院计算生物学与量化基因组学研究生学位，并于同年进入哈佛大学生物统计系攻读生物统计学博士学位。她的师从Prof. Peter Kraft，从事统计基因学相关研究，主要研究方向为利用统计学方法进行multi-omics 数据整合。
Transcriptome-wide association study (TWAS) test the association between imputed gene expression and trait. The power of the test depends largely on the accuracy of imputation weights and the association signal between the SNPs and phenotype of interest, thus bounded by the eQTL dataset used to build gene expression model and GWAS dataset. The power of TWAS could be affected when the eQTL or GWAS sample size is small. Here, we improve the power of TWAS through a cross-tissue approach for the eQTL dataset and multi-trait approach for the GWAS dataset. The two methodologies were shown to significantly improve the power of TWAS with simulation and real data applications. Here, we implement a pipeline to perform cross-cancer, cross-tissue TWAS analysis. We use newly-developed multi-trait TWAS test statistics to integrate the TWAS results for association between 11 separated cancers and predicted gene expression in each of 43 GTEx tissues; these include a "sum" test and a "variance components" test, analogous to fixed- and random-effects meta-analyses. We then integrated the results across different tissues using the Aggregated Cauchy Association Test (ACAT) combined test, a novel powerful and robust test for combining association results under general correlation patterns.