Polygenic risk scores (PRSs) are used to identify individuals at increased genetic risk of type 1 diabetes mellitus (T1DM). However, scores derived from European genetic datasets often show reduced predictive performance in individuals from non-European ancestries. A study published in Diabetologia developed and evaluated a trans-ancestry polygenic score to improve prediction across diverse populations.

The analysis used the PRS-CSx statistical framework and genome-wide association data from 29,469 T1DM cases across European, East Asian, African American, and Hispanic populations. The resulting trans-ancestry polygenic score (TA-PS) combined a non–human leukocyte antigen (HLA) component containing more than one million genetic variants with the HLA component of the previously reported European score GRS2x. Predictive performance was assessed using area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity in a multi-ancestry case-control cohort from Montreal, Canada (n = 4,657; non-European = 556). Findings were validated in four independent cohorts.

In the Montreal cohort, TA-PS achieved an AUROC of 0.89 compared with 0.85 for GRS2x. At the 90th percentile risk threshold, TA-PS sensitivity was 0.71 in Europeans and 0.77 in South Asians, while GRS2x sensitivity was 0.56 in Europeans and 0.32 in African Americans. Specificity with TA-PS remained ≥0.83 across ancestries. Validation in the CHOP-CAG, GRACE, All of Us, and UK Biobank cohorts showed consistent results. The trans-ancestry score demonstrated stronger discrimination and more consistent performance across populations than the European-derived score.