Multivariate Set Tests for Genetic Association Studies

In the last decade, genome-wide association studies (GWAS) have helped to advance our understanding of the genetic architecture of important traits, including human diseases. However, optimal analysis of GWAS data remains a major challenge. First, many traits of interest have polygenic architectures, which means that they are controlled by a large number of loci, which in turn can harbor multiple causal variants. Second, these loci tend to have effects on multiple related phenotypes, a phenomenon that is known as pleiotropy. Finally, genetic effects can depend on the specific cellular and environmental context. In this talk, I will present new statistical models to characterize polygeneicity, pleiotropy, and to elucidate context-specificity, by enabling joint genetic analysis across multiple variants, traits and contexts. First, I will discuss mtSet, an efficient mixed-model approach that enables genome-wide association testing between sets of genetic variants and multiple traits while accounting for confounding factors. Both in simulations and in applications to real datasets, we find that this joint modelling approach offers power advantages compared to methods that aggregate either across traits or variants in isolation. In the second part of the talk, I will discuss extensions to this model, deriving a new strategy to test for interactions between sets of genetic variants and categorical contexts (iSet). iSet accounts for polygenic effects and allows for characterizing context-specificity at genetic loci. In an application to a monocyte eQTL dataset, this method increases power for detecting genotype-context interactions and identifies genes associated with different regulatory architectures between contexts. Our results highlight that even “simple” traits such as gene expression levels can have surprisingly complex cis-genetic architectures, with multiple associated variants and changes of genetic regulation between different cellular contexts. First, I will discuss mtSet, an efficient mixed-model approach that enables genome-wide association testing between sets of genetic variants and multiple traits while accounting for confounding factors. Both in simulations and in applications to real datasets, we find that this joint modelling approach offers power advantages compared to methods that aggregate either across traits or variants in isolation. In the second part of the talk, I will discuss extensions to this model, deriving a new strategy to test for interactions between sets of genetic variants and categorical contexts (iSet). iSet accounts for polygenic effects and allows for characterizing context-specificity at genetic loci. In an application to a monocyte eQTL dataset, this method increases power for detecting genotype-context interactions and identifies genes associated with different regulatory architectures between contexts. Our results highlight that even “simple” traits such as gene expression levels can have surprisingly complex cis-genetic architectures, with multiple associated variants and changes of genetic regulation between different cellular contexts.

Date:
Speakers:
Francesco Paolo Casale
Affiliation:
University of Cambridge and the European Bioinformatics Institute