Development and validation of a Trans-Ancestry polygenic risk score for Type 1 Diabetes

ObjectivesThe high heritability of type 1 diabetes has enabled the development of polygenic risk scores (PRS) as disease risk screening tools. PRS can identify individuals at the highest genetic risk in a population, who can benefit from autoantibody and metabolic surveillance, to avoid ketoacidosis at diagnosis and access preventive therapies. However, PRS for type 1 diabetes developed from European data perform less well in non-European ancestries. We aimed to develop a PRS with comparable performance among different ancestries. MethodsUsing a the PRS-CSx method, and data from large European, East-Asian, African-American and Hispanic type 1 diabetes GWAS (Ntotal_cases=29,469), we developed a trans-ancestry PRS (TA-PS), combining a non-HLA component incorporating over a million variants, with the HLA component of a published European PRS (GRS2x). We tested the performance of the PRS using AUROC, sensitivity and specificity in a multi-ancestry T1D case-control cohort (Ntotal= 4,657; Nnon-European=556) from Montreal, Canada. We validated our results in two independent T1D case-control cohorts (CHOP-CAG and GRACE) and two population-based cohorts (All of Us and UK Biobank). ResultsIn our multi-ancestry Montreal-based cohort, TA-PS showed an AUROC of 0.89 which was significantly higher from the AUROC of 0.85 of GRS2x. At a 90th percentile cut-off, in African-Americans, the sensitivity of GRS2x was 0.32, compared to 0.56 in Europeans. For TA-PS, we obtained overall better sensitivities, ranging from 0.71 in Europeans to 0.77 in South Asians. TA-PS demonstrated slightly lower albeit acceptable specificity compared to that of GRS2x (> 0.83 across all ancestries). These results were validated in the four independent cohorts. ConclusionWe developed a trans-ancestry PRS that outperformed the European-based GRS2x. Importantly, TA-PS provides a comparable prediction in various ancestries, which supports its use in population-wide screening programs. Research in contextO_ST_ABSWhat is already known about this subject?C_ST_ABS- Polygenic risk scores (PRS) for type 1 diabetes are primarily developed using data from individuals of European ancestry. - The widely used, European-based GRS2 score shows reduced performance in non-European populations, particularly among individuals of African descent. - There are concerns regarding the equity of genetic risk prediction in population-based screening programs for T1D. What is the key question?- Can a trans-ancestry PRS provide accurate and equitable type 1 diabetes risk prediction across ancestries? What are the new findings?- A new trans-ancestry score, TA-PS, was developed by integrating an optimized non-HLA PRS to GRS2. - Compared to GRS2, in multi ancestry case-control and population-based cohorts, TA-PS improves sensitivity across all ancestry groups while maintaining high specificity. How might this impact on clinical practice in the foreseeable future?- TA-PS could provide equitable genetic risk stratification in population-wide screening programs for type 1 diabetes.