TRIDENT: a framework for robust multi-trait GWAS identifies 66 novel multi-trait osteoarthritis signals

Abstract

As multi-trait genome-wide association studies (GWAS) are increasingly used to identify shared genetic associations across related phenotypes, practical approaches to assess the robustness of their findings are lacking. Here we present a three-step framework (Trident) for robust multi-trait GWAS that uses an earlier, smaller GWAS meta-analysis to test whether phenotypes can be validly combined as well as the latest, largest GWAS meta-analysis of the same phenotypes for discovery, followed by translational annotation to assess disease relevance and prioritize likely effector genes. We applied Trident by using the Combined-GWAS (C-GWAS) method to osteoarthritis, a degenerative joint disease, across five osteoarthritis joint sites. Signals identified in the earlier GWAS meta-analysis showed high validation in the replication dataset, supporting the robustness of this approach. Applied to the latest and largest osteoarthritis GWAS meta-analysis, C-GWAS identified 66 novel associations not identified with conventional single-trait GWAS meta-analyses, including signals with shared and discordant effects across different joint sites. Translational annotation linked these signals to biologically plausible osteoarthritis genes and pathways. Together, we provide a practical framework for robust multi-trait GWAS that increases detection power by identifying novel signals and, by applying it to the example of osteoarthritis of five joints, refine the genetic architecture of this common disease.