Multi-ancestry polygenic risk scores for the prediction of type 2 diabetes and complications in diverse ancestries
Bibliographic record
Abstract
BACKGROUND: Polygenic risk scores (PRSs) improve prediction of the development of type 2 diabetes over the use of clinical risk factors alone; however, they perform poorly in populations of non-European ancestry, limiting their global clinical utility. We aimed to deliver comprehensive and rigorously tested multi-ancestry PRSs for prediction in type 2 diabetes. METHODS: We conducted meta-analyses using data from type 2 diabetes genome-wide association studies (GWAS) across cohorts from five major global ancestries: European, African or African American, Admixed American, South Asian, and East Asian. We used summary statistics from the GWAS to construct single-ancestry PRSs (using the continuous-shrinkage PRS-CS method) and multi-ancestry PRSs (using the PRS-CSx method), and constructed ancestry-specific linkage disequilibrium panels to model pairwise correlations between single-nucleotide polymorphisms in GWAS during PRS construction. Models were validated for association with type 2 diabetes in at least four independent cohorts per ancestry. The effect sizes of PRSs were estimated as the odds ratio (OR) per SD of the PRS, and ORs for individuals at the 90th, 95th, and 97·5th PRS percentiles were compared with the IQR as a reference. We also tested our PRS models for prediction of diabetes incidence with or without additional clinical factors, as well as microvascular complications and comorbidities. FINDINGS: Our analysis used data from 409 959 individuals with type 2 diabetes and 1 983 345 controls: respectively, 359 819 and 1 825 729 indivduals were included in the GWAS dataset, with 10 992 and 31 792 individuals in the training dataset and 39 148 and 125 824 individuals in the validation dataset. The best predictive performance for the single-ancestry PRSs was in European (incremental AUC 0·07-0·14) and East Asian (0·02-0·16) ancestries, whereas prediction was poorer for African or African American (0·02-0·03), Admixed American (0·02-0·04), and South Asian (0·02-0·04) ancestries, correlating with sample sizes in the GWAS. Compared with single-ancestry PRSs, our multi-ancestry PRSs showed higher effect sizes and smaller 95% CIs across all ancestries: OR per SD 1·73 (95% CI 1·67-1·80) in African or African American, 2·82 (2·67-2·97) in Admixed American, 2·45 (2·36-2·54) in East Asian, 2·36 (2·32-2·41) in European, and 2·23 (2·05-2·42) in South Asian ancestries. Individuals in the 97·5th PRS percentile had a 3-7 times increased risk of type 2 diabetes compared with those in the IQR (OR 3·43 [95% CI 2·80-4·21] in African or African American, 7·47 [5·64-9·89] in Admixed American, 6·62 [5·58-7·85] in East Asian, 6·25 [5·72-6·82] in European, and 4·50 [2·70-7·53] in South Asian ancestries). These PRSs were also associated with earlier onset of type 2 diabetes, higher risk of developing microvascular complications, and provide additional predictive value beyond clinical factors. In individuals with type 2 diabetes, the association between multi-ancestry PRSs and risk of microvascular complications and comorbidity was studied in populations of African, Admixed American, and European ancestries and was significant in all three ancestry groups for diabetic retinopathy (ORs per SD 1·28-1·57), diabetic nephropathy (1·25-1·58), proliferative diabetic retinopathy (1·39-2·08), and end-stage diabetic nephropathy (1·44-1·87); PRS was associated with coronary artery disease in the Admixed American ancestry group only (1·16 [95% CI 1·08-1·25]). INTERPRETATION: These validated, publicly available PRSs can improve risk stratification for type 2 diabetes onset and complications across diverse ancestries, supporting their further evaluation in clinical settings. FUNDING: The National Human Genome Research Institute of the US National Institutes of Health.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.023 | 0.029 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.006 |
| Bibliometrics | 0.003 | 0.004 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".