Novel joint enrichment test demonstrates high performance in simulations and identifies cell-types with enriched expression of inflammatory bowel disease risk loci

Abstract

A number of methods have been developed to assess the enrichment of polygenic risk variants - from summary statistics of genome-wide association studies (GWAS) - within specific gene-sets, pathways, or cell-type signatures. The assumptions made by these methods vary, which leads to differences in results and performance across different genetic trait architectures and sample sizes. We devise a novel statistical test that combines independent signals from each of three commonly-used enrichment tests (LDSC, MAGMA & SNPsea) into a single P-value, called the block jackknife GWAS joint enrichment test (GWASJET). Through simulations, we show that this method has comparable or greater power than competing methods across a range of sample sizes and trait architectures. We use our new test in an extensive analysis of the cell-type specific enrichment of genetic risk for inflammatory bowel disease (IBD), including Crohns disease (CD) and ulcerative colitis (UC). Counterintuitively, we find stronger enrichments of IBD risk genes in older gene expression data from bulk immune cell-types than in single-cell data from inflamed patient intestinal samples. We demonstrate that GWASJET removes many seemingly-spurious enriched cell-types identified by other methods, and identifies a core set of immune cells that express IBD risk genes, particularly myeloid cells that have been experimentally stimulated. We also demonstrate that many cell-types are differentially enriched for CD compared to UC risk genes, for example gamma-delta T cells show stronger enrichment for CD than UC risk genes. Author summaryGenetic association studies have discovered a number of DNA variations that are associated with heritable human diseases and traits. One method of investigating the functions of these variants is to test whether they are enriched in parts of the genome associated with specific cell-types or cell conditions - defined by gene expression data or other similar data types. However, there are a number of published statistical methods to test such enrichments; these methdos make different assumptions and their results can vary, sometimes dramatically. We present a novel consensus method, called GWASJET, that combines the results of these different methods to produce a single result. We show that GWASJET can outperform individual methods in simulations. We apply this method to gene expression data from a number of tissues and conditions relevant to inflammatory bowel diseases (IBD). Our method removes potentially false results based on a priori biological knowledge, and reveals that IBD genes are generally clustered in a large number of immune cell-types, especially myeloid cells treated with specific stimulatory molecules.