- This event has passed.
Webinar: Biostatistics seminar – “Testing 15 Billion eQTL Associations: Fine-Mapping and Multivariate Analysis Methods”
November 11, 2020, 12:00 pm to 1:00 pm
“Testing 15 Billion eQTL Associations: Fine-Mapping and Multivariate Analysis Methods” will be presented by Gao Wang, PhD, Assistant Professor of Neurological Sciences, Department of Neurology, Columbia University.
In recent years, genetic association analysis in human genetics are typically performed for millions of variables (SNPs) over thousands of phenotypes, ranging from complex diseases to molecular phenotypes. Due to the presence of linkage disequilibrium (LD), localizing non-zero effect SNP from many other correlated ones is a hard problem. Particularly when SNPs have non-zero effect in multiple conditions, it is both interesting and challenging to understand whether it is truly pleiotropic effect, or artifact due to LD.
In this presentation, Dr. Wang introduces a new Bayesian regression model for genetic fine-mapping, which his group calls the Sum of Single Effects (SuSiE) model. They also introduce a corresponding new fitting procedure for SuSiE, called Iterative Bayesian Step-wise Selection (IBSS). IBSS algorithm computes a variational approximation to the posterior distribution under the SuSiE model, which is significantly faster than state-of-the-art fine-mapping methods that uses other approaches to approximate posterior. The model structure also yields, by design, independent confidence sets, each designed to capture one association signal, making the results easy to interpret and ideal for guiding follow-up studies. When extended to performing multivariate analysis, the researchers developed mvSuSiE with a multivariate adaptive shrinkage prior, which leverages sharing of genetic effects in data through a rigorous empirical Bayes framework. The team applied mvSuSiE to cis-eQTL analysis in GTEx data and showed that mvSuSiE elucidates genetic architecture of molecular phenotypes across human tissues, and has the potential to be applied to jointly analyze many GWAS traits.
Participants may also join by calling 929-205-6099, meeting ID 322 043 373. If prompted, enter passcode 167660.