Genome-Wide Polygenic Risk Scores and prediction of Gestational Diabetes in South Asian Women

Gestational diabetes Mellitus (GDM) affects 1 in 7 births and is associated with numerous adverse health outcomes for both mother and child. GDM is suspected to share a large common genetic background with type 2 diabetes (T2D). In this study, we first sought to build and characterize different GDM genome-wide polygenic risk scores (PRSs) using whole genome genotypes from the South Asian Birth Cohort (START) and data from the DIAGRAM consortium. The second aim of this study was to estimate the heritability of GDM. GDM PRSs were derived for 832 South Asian women participating in the START study using three methods (Pruning and thresholding (P+T), LDpred, and the gradient boosted and LD adjusted (GraBLD) methods). Summary statistics were derived from Mahajan et al., 2014 and Scott et al., 2017, two large genome-wide association meta-analysis performed in ethnically diverse and European participants respectively. Linkage disequilibrium (LD) between variants was assessed using the START and 1000 Genomes (1KG) data. Both weighted and unweighted PRSs were derived. Association with GDM was tested using logistic regression. Heritability of GDM was estimated using the GRMEL approach. Results were replicated in South Asian and European women from the UK Biobank (UKB) study. The best P+T GraBLD and LDpred PRSs were all based on data from Mahajan et al., but differed with respect to their source of LD. The best PRS was highly associated with incident GDM in START (AUC= 0.62, OR: 1.60 [95%CI: 1.44 - 1.69]) and in South Asian (AUC=0.65) and British (AUC=0.58) women from UKB. Heritability of GDM approximated 0.83 [95%CI: 0.38 - 1] in START and 0.17 [95%CI: 0.03 - 0.31] in South Asian women from UKB. Our results highlight the importance of combining whole genome data, summary statistics from large multi-ethnic genome-wide meta-analysis in order to derive an optimal PRSs and to estimate the heritability of GDM in South Asian women.