Leveraging gene co-expression patterns to infer trait-relevant tissues in genome-wide association studies.

Genome-wide association studies (GWASs) have identified many SNPs associated with various common diseases. Understanding the biological functions of these identified SNP associations requires identifying disease/trait relevant tissues or cell types. Here, we develop a network method, CoCoNet, to facilitate the identification of trait-relevant tissues or cell types. Different from existing approaches, CoCoNet incorporates tissue-specific gene co-expression networks constructed from either bulk or single cell RNA sequencing (RNAseq) studies with GWAS data for trait-tissue inference. In particular, CoCoNet relies on a covariance regression network model to express gene-level effect measurements for the given GWAS trait as a function of the tissue-specific co-expression adjacency matrix. With a composite likelihood-based inference algorithm, CoCoNet is scalable to tens of thousands of genes. We validate the performance of CoCoNet through extensive simulations. We apply CoCoNet for an in-depth analysis of four neurological disorders and four autoimmune diseases, where we integrate the corresponding GWASs with bulk RNAseq data from 38 tissues and single cell RNAseq data from 10 cell types. In the real data applications, we show how CoCoNet can help identify specific glial cell types relevant for neurological disorders and identify disease-targeted colon tissues as relevant for autoimmune diseases. Author summary: Identifying trait-relevant tissues or cell types is important for understanding disease etiology. Several computational methods have been recently developed to integrate omics studies with genome-wide association studies (GWASs) in order to infer trait-relevant tissues or cell types. For example, both LDSC-SEG and RolyPoly rely on genes that are specifically expressed in a given tissue or cell type for inferring trait-tissue relevance. However, these previous methods have thus far ignored an important biological feature of gene expression data; that is, genes are interconnected with each other and are co-regulated together. Such gene co-expression pattern occurs in a tissue specific or cell type specific fashion and may contain invaluable information for inferring trait-tissue relevance. Here, we develop a network model to take advantage of the tissue-specific or cell type specific gene co-expression patterns inferred from bulk RNA sequencing or single cell RNA sequencing studies into GWASs. We illustrate the benefits of our method in identifying trait-relevant tissues or cell types through simulations and applications to real data sets. [ABSTRACT FROM AUTHOR]