Genome-wide polygenic risk score for type 2 diabetes in Indian population

Abstract

Genome-wide polygenic risk scores (PRS) for lifestyle disorders, like Type 2 Diabetes (T2D), are useful in identifying at-risk individuals early on in life, and to guide them towards healthier lifestyles. The current study was aimed at developing PRS for the Indian population using imputed genotype data from UK Biobank and testing the developed PRS on data from GenomegaDB of Indians living in India. 959 T2D cases and 2,818 controls were selected from Indian participants of UK Biobank to develop the PRS. Summary statistics available for South Asians, from the DIAMANTE consortium, were used to weigh genetic variants. LDpred2 algorithm was used to adjust the effect of linkage disequilibrium among the variants. The association of PRS with T2D, after adjusting for age, sex and top ten genetic principal components, was found to be very significant (AUC = 0.7953, OR = 2.9856 [95% CI: 2.7044–3.2961]). When participants were divided into four PRS quartile groups, the odds of developing T2D increased sequentially with the higher PRS groups. The highest PRS group (top 25%) showed 5.79 fold increased risk compared to the rest of the participants (75%). The PRS derived using the same set of variants was found to be significantly associated with T2D in the test dataset of 445 Indians (AUC = 0.7781, OR = 1.6656 [95%CI = 0.6127–4.5278]). Our study demonstrates a framework to derive Indian-specific PRS for T2D. The accuracy of the derived PRS shows it’s potential to be used as a prognostic metric to stratify individuals, and to recommend personalized preventive strategies.

Genome‐wide association and polygenic risk score estimation of type 2 diabetes mellitus in Kinh Vietnamese—A pilot study

“…The capability of T2DM prediction could be increased by adding clinical data to the model. 32 , 33 Ideally, these data should be collected at the time of diagnosis and prior to any treatment. Given the nature of a cross‐sectional study, we could not obtain clinical parameters at the time of diagnosis for all the T2DM cases; therefore, combining clinical parameters in the model to predict T2DM was not appropriate.…”

“…The best predictive model for T2DM using only genetic data in this study has AUC = 0.70. The capability of T2DM prediction could be increased by adding clinical data to the model 32,33 . Ideally, these data should be collected at the time of diagnosis and prior to any treatment.…”