Abstract

Genome-wide association studies (GWAS) look for associations between millions of genomic variants and traits such as disease or treatment response, and have revealed shared genetic architectures across disease collections, such as the immune-mediated diseases (IMD) which include type 1 diabetes, asthma, or rheumatoid arthritis. However, integrative approaches to describe these architectures beyond pairwise comparison have been limited by the dimension of these statistics, as well as different sharing patterns across diseases, rendering generalisations about inter-disease relationships difficult. Here, I describe a new method based on principal component analysis and Bayesian shrinkage to engineer genetic risk components from GWAS summary statistics for a set of related traits. This enables us to detect significant associations in the limited sample numbers available from less frequent diseases, which would otherwise have insufficient powered for genetic study. I also show that using potential mediating traits, such as blood cell counts in the engineering can help identify more interpretable components, and thus help understand the genetic architecture of IMD and infectious diseases, among others.