It remains both important and challenging to boost statistical power for GWAS to identify more trait-associated SNPs or loci to account for ``missing heritability". Furthermore, since most GWAS trait-associated SNPs are not in gene coding regions, a biological interpretation of their function is largely missing. To overcome the above two shortcomings of conventional GWAS, an emerging theme of many recent approaches is to integrate GWAS with expression QTL (eQTL), methylation QTL (meQTL), enhancer-promoter interaction (EPI)or other molecular QTL and omic data, due to their regulatory roles and their enrichment with GWAS trait-associated SNPs. We first introduce a powerful and adaptive gene-based testing framework, then illustrate its application to integrating schizophrenia GWAS summary data with meQTL and EPI annotations. We demonstrate that the proposed method could identify some significant and novel genes (containing no genome-wide significant SNPs nearby) that would be missed by other existing approaches, including the standard and some integrative gene-based association methods such as PrediXcan/TWAS. Finally, we briefly discuss a convolutional neural network (CNN) approach, coupled with transfer learning, to predicting cell line-specific EPIs based on DNA sequence data. Light refreshments for seminar guests will be served at 3:10 p.m. in TBD
Department of BiostatisticsAdaptive gene-based testing approaches to integrating GWAS with omic data
Wei Pan, Ph.D. - School of Public Health, University of Minnesota
December 6, 2018
3:30 pm - 5:00 pm
3755 SPH I
1415 Washington Heights
Ann Arbor, MI 48109-2029
Sponsored by: Department of Biostatistics
Contact Information: Zhenke Wu (zhnekewu@umich.edu), Peisong Han (peisong@umich.edu)
It remains both important and challenging to boost statistical power for GWAS to identify more trait-associated SNPs or loci to account for ``missing heritability". Furthermore, since most GWAS trait-associated SNPs are not in gene coding regions, a biological interpretation of their function is largely missing. To overcome the above two shortcomings of conventional GWAS, an emerging theme of many recent approaches is to integrate GWAS with expression QTL (eQTL), methylation QTL (meQTL), enhancer-promoter interaction (EPI)or other molecular QTL and omic data, due to their regulatory roles and their enrichment with GWAS trait-associated SNPs. We first introduce a powerful and adaptive gene-based testing framework, then illustrate its application to integrating schizophrenia GWAS summary data with meQTL and EPI annotations. We demonstrate that the proposed method could identify some significant and novel genes (containing no genome-wide significant SNPs nearby) that would be missed by other existing approaches, including the standard and some integrative gene-based association methods such as PrediXcan/TWAS. Finally, we briefly discuss a convolutional neural network (CNN) approach, coupled with transfer learning, to predicting cell line-specific EPIs based on DNA sequence data. Light refreshments for seminar guests will be served at 3:10 p.m. in TBD