The T1D MAPS polygenic score significantly improved type 1 diabetes risk prediction compared to T1D GRS2 in non-European populations (ΔAUC=0.052, p=2.9x10-6).
Observational
Yes
Does T1D MAPS improve the prediction of Type 1 Diabetes risk compared to T1D GRS2 across diverse ancestral populations?
T1D MAPS provides a statistically significant improvement over T1D GRS2 for predicting Type 1 Diabetes risk in non-European populations while maintaining similar performance in European populations.
Effect estimate: ΔAUC 0.052
p-value: p=2.9x10-6
Introduction and Objective: Polygenic scores (PS) are powerful predictors of type 1 diabetes (T1D) risk, but their performance can differ substantially across ancestries, highlighting the need for rigorous evaluation in different populations. In this study, we evaluated the performance of several established PS to predict T1D risk across ancestries in four biobanks: All of Us v8 (AoU), Genomic Health Initiative at Endeavor Health (GHI), Penn Medicine BioBank (PMBB), and Geisinger MyCode. Methods: We assessed several T1D PS - including T1D GRS2, T1D GRS2’, T1GRS, T1D MAPS, and TA-PS - using data from the AoU v8 cohort. We identified individuals with and without T1D based on the eMERGE T1D algorithm. The area under the receiver operating characteristic curve (AUC) was estimated separately for individuals with European (EUR) and non-European (non-EUR) genetic ancestry, and differences between scores (ΔAUC) were assessed using DeLong’s test. We further performed meta-analyses of ΔAUC between T1D GRS2 and T1D MAPS across AoU, GHI, PMBB, and MyCode, evaluating EUR and non-EUR groups separately. Results: We analyzed 868 individuals with T1D in AoU (563 EUR), 160 in GHI (113 EUR), 290 in PMBB (199 EUR), and 1082 in MyCode (922 EUR). In AoU, among the PS we tested, T1D MAPS had the highest AUC across all ancestries (AUCEUR=0.87, AUCnon-EUR=0.72). In EUR, T1D MAPS and T1D GRS2 had similar performance (ΔAUC=0.002, p=0.23), whereas in non-EUR populations, T1D MAPS outperformed T1D GRS2 (ΔAUC=0.072, p=0.0048). In a meta-analysis across four biobanks, T1D MAPS showed similar performance to T1D GRS2 in EUR ancestry (ΔAUC=0.016, p=0.33) and significantly increased performance in non-EUR populations (ΔAUC=0.052, p=2.9x10-6). Conclusion: T1D MAPS provides a consistent and statistically significant improvement over T1D GRS2 in non-EUR populations and performs similarly to T1D GRS2 in EUR individuals. These results highlight the value of incorporating diverse ancestries into T1D genetic risk model development and evaluation. Disclosure S. Nam: None. J. Li: None. R. Mandla: None. H. Tran: None. K. Taylor: None. M. Vora: None. A. Huerta: None. J. Wei: None. A. Mulford: None. A.R. Sanders: None. J. Xu: Advisory Panel; Current; GenomicMD, GoPath Labs. L.K. Billings: Advisory Panel; Current; Novo Nordisk, Lilly, Sanofi, Amgen Inc., Bayer AG. B. Pasaniuc: None. D. Carey: None. U. Mirshahi: None. J.C. Florez: Research Support; Current; Novo Nordisk. Consultant; Current; Alveus Therapeutics. A. Manning: None. J. Mercader: None. M. Udler: Advisory Panel; Ended; Novo Nordisk. Research Support; Current; Novo Nordisk. A.J. Deutsch: None. Funding National Institutes of Health (NIH) / National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) K23 DK140643
Nam et al. (Sun,) conducted a observational in Type 1 Diabetes. T1D MAPS (Type 1 Diabetes Multiancestry Polygenic Score) vs. T1D GRS2 was evaluated on Difference in Area under the receiver operating characteristic curve (ΔAUC) for T1D risk prediction (ΔAUC 0.052, p=2.9x10-6). The T1D MAPS polygenic score significantly improved type 1 diabetes risk prediction compared to T1D GRS2 in non-European populations (ΔAUC=0.052, p=2.9x10-6).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: