Integration of Expression QTLs with fine mapping via SuSiE

Abstract Genome-wide association studies (GWASs) have achieved remarkable success in associating thousands of genetic variants with complex traits. However, the presence of linkage disequilibrium (LD) makes it challenging to identify the causal variants. To address this critical gap from association to causation, many fine mapping methods have been proposed to assign well-calibrated probabilities of causality to candidate variants, taking into account the underlying LD pattern. In this manuscript, we introduce a statistical framework that incorporates expression quantitative trait locus (eQTL) information to fine mapping, built on the sum of single-effects (SuSiE) regression model. Our new method, SuSiE 2 , connects two SuSiE models, one for eQTL analysis and one for genetic fine mapping. This is achieved by first computing the posterior inclusion probabilities (PIPs) from an eQTL-based SuSiE model with the expression level of the candidate gene as the phenotype. These calculated PIPs are then utilized as prior inclusion probabilities for risk variants in another SuSiE model for the trait of interest. By leveraging eQTL information, SuSiE 2 enhances the power of detecting causal SNPs while reducing false positives and the average size of credible sets by prioritizing functional variants within the candidate region. The advantages of SuSiE 2 over SuSiE are demonstrated by simulations and an application to a single-cell epigenomic study for Alzheimer’s disease. We also demonstrate that eQTL information can be used by SuSiE 2 to compensate for the power loss because of an inaccurate LD matrix. Author summary Genome-wide association studies (GWASs) have proven powerful in detecting genetic variants associated with complex traits. However, there are challenges in distinguishing the causal variants from other variants strongly correlated with them. To better identify causal SNPs, many fine mapping methods have been proposed to assign well-calibrated probabilities of causality to candidate variants. We introduce a statistical framework that incorporates expression quantitative trait locus (eQTL) information to fine mapping, which can improve the accuracy and efficiency of association studies by prioritizing functional variants within the risk genes before evaluating the causation. Our new fine mapping framework, SuSiE 2 , connects two sum of single-effects (SuSiE) models, one for eQTL analysis and one for genetic fine mapping. The posterior inclusion probabilities from an eQTL-based SuSiE model are utilized as prior inclusion probabilities for risk variants in another SuSiE model for the trait of interest. Through simulations and a real data analysis focused on Alzheimer’s disease, we demonstrate that SuSiE 2 improves fine mapping results by simultaneously increasing statistical power, controlling the type I error rate, and reducing the average size of credible sets.