The Genetic Architecture of Diabetic Retinopathy
Project Summary/Abstract
Diabetic retinopathy (DR) is the leading cause of vision loss and preventable blindness in adults, afflicting an
estimated 93 million individuals worldwide. As the global prevalence of diabetes mellitus (DM) rises, the global
prevalence of diabetic retinopathy will also rise. The eye is highly susceptible to damage from DM due to the
delicate structures and intricate control of homeostasis in the ocular environment. DR has long been
recognized as a microvascular disease triggered by the breakdown of the blood-retinal barrier and
neovascularization in the retina. The course of diabetic disease and severity of sequelae vary substantially
between cases, and the cause for this variability is not explained well by known risk factors. Several genome-
wide association studies (GWAS) of DR have been completed, and while a modest number of significant loci
have been reported only a single locus has been significantly associated with DR after a replication. These
studies were conducted in collaborative consortia of cohort studies and the largest study to date included
22,279 cases and 23,977 diabetic controls. We will leverage the previous studies in combination with
resources from the Million Veteran Program (MVP), BioVU, and the eMERGE Network (an estimated 77,518
cases, 133,295 type II diabetic controls) to test the hypothesis that common genetic variants are associated
with DR in the largest multi-ethnic GWAS to date. AIM 1. We will combine evidence for association from
logistic regression analysis of DR-SNP relationships across the resources described above using inverse
variance-weighted fixed effects meta-analyses, both within and across racial groups. The majority of previous
studies focused on detecting SNP-phenotype associations and did not evaluate regulatory genetic effects or
polygenic and causal effects for DR. AIM 2. We will use the summary statistics from the meta-analysis and the
Gene-Tissue Expression Project in an analysis using S-PrediXcan to evaluate associations between DR and
genetically predicted gene expression. Additionally, we will identify genetically predicted gene effects that are
not due to linkage disequilibrium contamination through the use of a colocalization analysis. Colocalization
methods are under constant development, and we will monitor the literature for best practices. AIM 3. We will
develop a genome-informed predictive model, combining a diabetic retinopathy polygenic risk score (PRS), DR
PheWAS results, and current clinical risk factors for DR. We will develop a DR PRS utilizing the summary
statistics of the MVP GWAS and Pollack et al., followed by training in BioVU. A PheWAS analysis incorporating
the DR PRS can elucidate previously unknown association between DR associated loci and other disease
traits. We will combine the PRS, PheWAS results, and current clinical risk factors into the genome informed
predictive model that will be validated in the eMERGE Network. The outlined study will significantly increase
the sample size of previous studies of DR, evaluate regulatory effects on DR, and develop predictive models
that may have utility in clinical decision making of DR patients.