Abstract
A Type 1 Diabetes Genetic Risk Score (T1DGRS) aids diagnosis and prediction of Type 1 Diabetes (T1D). While traditionally derived from imputed array genotypes, Whole Genome Sequencing (WGS) provides a more direct approach and is used increasingly in research studies. We aim to assess differences between WGS-based T1DGRS and array-based T1DGRS, focusing on variations across genetic ancestries. We generated 67-variant T1DGRS from 149,265 individuals from UK Biobank with WGS, TOPMed-imputed, and 1000 Genomes-imputed array genotypes. WGS-based T1DGRS showed strong correlation to GRS from TOPMed-imputed array genotypes (r = 0.99), with a slightly lower mean (-0.0028 SD, p < 10− 31). Correlation was lower in both non-European populations and GRS from 1000 Genomes-imputed array genotypes (r ranging between 0.95–0.98). This can lead to between 6–29% re-categorisation of individuals at clinical risk thresholds using the array-based GRS in non-European populations. Compared to Europeans, WGS-based T1DGRS was much lower for African and South Asian populations. In conclusion, WGS is a viable approach for generating T1DGRS and TOPMed-imputed genotypes offer a cost-effective alternative. The observed variations in T1DGRS at the population-level among different genetic ancestries cautions against indiscriminate use of European-centric T1DGRS risk thresholds in clinical practice and advocates the need for ancestry-specific or pan-ancestry standards.