Genome-wide polygenic risk score method for diabetic kidney disease in patients with type 2 diabetes

Diabetic kidney disease (DKD) is considered partially hereditary, but the genetic factors underlying disease remain largely unknown. A key barrier to our understanding stems from its heterogeneity, and likely polygenic etiology. Proteinuric and non-proteinuric DKD are two sub-classes of DKD, defined by high urinary albumin-to-creatinine ratio (UACR) and low creatinine estimated glomerular filtration rate (eGFR). Prior genome-wide association studies (GWAS) have identified multiple loci associated with eGFR and UACR. We aimed to combine summary statistics from previous GWAS for eGFR and UACR in one prediction model and associate it with DKD prevalence. We then tested this using genetic data from 18,841 individuals diagnosed with type 2 diabetes in UK Biobank. We computed two genome-wide polygenic risk scores (GPS) aggregating effects of common variants associated with the two measurements, eGFR and UACR. We show that including both GPS in a single model confers significant improvement in comparison with the single GPS model generated from GWAS summary statistics for DKD. We also find in replication analysis in 5,389 individuals in the multi-ethnic BioMe Biobank, that although the combined model had consistent direction of association, the lowest performance was in individuals with recent African ancestry. In summary, we show that joint modeling of polygenic associations of eGFR and UACR is more significantly associated with DKD than individual modeling as well as a GPS comprised of only DKD summary statistics and may be used to gain insights into biology and progression. However, efforts should be made to develop and validate polygenic approaches in diverse populations.