Orienting the causal direction between two variables
University of Michigan School of Public Health
1755 SPH I, 1415 Washington Heights Ann Arbor, MI 48109-2029
Causal inference, going beyond typical (GWAS) association analyses, is crucial for unraveling the cause of disease, which, if successful, will undoubtedly facilitate developing effective treatment and prevention strategies. However, it is well-known that causal inference is extremely challenging with data from only observational studies. Taking advantage of existing GWAS and their identified SNPs associated with complex traits, two-sample Mendelian randomization (MR) and its variant (2SLS), as implemented in genetics and called (generalized) summary data-based MR (SMR or GSMR) and transcriptome-wide association studies (TWAS or PrediXcan) respectively, are potentially useful in identifying causal risk factors for an outcome variable/trait. In practice, however, most existing methods, including both SMR and TWAS, assume a known causal direction between two variables/traits of interest (thus called the exposure/risk factor and the outcome respectively). A simple approach in the framework of MR using a single SNP as the instrumental variable has recently appeared in the literature. To improve its statistical efficiency, we extend the method to multiple SNPs in one or more loci. We will use some real GWAS examples to demonstrate the performance of the proposed method and compare it with several other existing methods. This is joint work with Haoran Xue. Light refreshments for seminar guests will be served at 3:10 p.m. in 1755 Department of Biostatistics

Orienting the causal direction between two variables

Wei Pan, PhD., Professor - Division of Biostatistics, University of Minnesota

icon to add this event to your google calendarFebruary 26, 2019
3:30 pm - 5:00 pm
1755 SPH I
1415 Washington Heights
Ann Arbor, MI 48109-2029
Sponsored by: Department of Biostatistics
Contact Information: Zhenke Wu (zhenkewu@umich.edu) & Peisong Han (peisong@umich.edu)

Causal inference, going beyond typical (GWAS) association analyses, is crucial for unraveling the cause of disease, which, if successful, will undoubtedly facilitate developing effective treatment and prevention strategies. However, it is well-known that causal inference is extremely challenging with data from only observational studies. Taking advantage of existing GWAS and their identified SNPs associated with complex traits, two-sample Mendelian randomization (MR) and its variant (2SLS), as implemented in genetics and called (generalized) summary data-based MR (SMR or GSMR) and transcriptome-wide association studies (TWAS or PrediXcan) respectively, are potentially useful in identifying causal risk factors for an outcome variable/trait. In practice, however, most existing methods, including both SMR and TWAS, assume a known causal direction between two variables/traits of interest (thus called the exposure/risk factor and the outcome respectively). A simple approach in the framework of MR using a single SNP as the instrumental variable has recently appeared in the literature. To improve its statistical efficiency, we extend the method to multiple SNPs in one or more loci. We will use some real GWAS examples to demonstrate the performance of the proposed method and compare it with several other existing methods. This is joint work with Haoran Xue. Light refreshments for seminar guests will be served at 3:10 p.m. in 1755