Type 2 diabetes polygenic risk score demonstrates context-dependent effects and associations with type 2 diabetes-related risk factors and complications across diverse populations

Polygenic risk scores (PRS) hold prognostic value for identifying individuals at higher risk of type 2 diabetes (T2D). However, further characterization is needed to understand the generalizability of T2D PRS in diverse populations across various contexts. We characterized a multi-ancestry T2D PRS among 244,637 cases and 637,891 controls across eight populations from the Population Architecture Genomics and Epidemiology (PAGE) Study and 13 additional biobanks and cohorts. PRS performance was context dependent, with better performance in those who were younger, male, with a family history of T2D, without hypertension, and not obese or overweight. Additionally, the PRS was associated with various diabetes-related cardiometabolic traits and T2D complications, suggesting its utility for stratifying risk of complications and identifying shared genetic architecture between T2D and other diseases. These findings highlight the need to account for context when evaluating PRS as a tool for T2D risk prognostication and potentially generalizable associations of T2D PRS with diabetes-related traits despite differential performance in T2D prediction across diverse populations.Here, we provide the T2D PRS PheWAS results meta-analyzed across all populations and by population, and PheWAS results in each of the five biobanks: the Icahn School of Medicine at Mount Sinai BioMe biobank in New York City (BioMe), BioVU Biobank (BioVU), Colorado Center for Personalized Medicine Biobank (CCPM), Million Veteran Program (MVP), and Michigan Genomics Initiative (MGI). The MVP PheWAS results were restricted to phecodes with more than 500 cases, and two PheCodes—HIV disease (071 and 071.1) and sickle cell anemia (282.50)—were removed.Utilizing the PheWAS R package, we ran logistic regression models to assess PRS-phenotype associations, adjusting for age, sex, BMI, and the first ten PCs. For the mapping of the International Classification of Diseases (ICD) 9 and ICD-10 codes to phecodes, we utilized phecodes version 1.2 for ICD-9-CM and the 2018 beta version of ICD-10-CM. Cases were considered valid if they had a minimum phecode count of two. Only phecodes with at least 10 cases were considered. A total of 1,815 unique phenotypes across 17 disease categories were tested across the five biobanks, with 1,777, 1,171, 1,813, and 1,695 phenotypes available in AFR, ASN, EUR, and HIS populations, respectively. Any values smaller than the smallest non-zero value that R can represent are denoted as 5.0E-324.For questions, please contact Boya Guo ([email protected]) or Burcu F. Darst ([email protected]).