New Approach Uses Public Genomics Data to Predict Risk of Cardiometabolic Disease in Non-European Populations
Key findings
- This study leveraged publicly available genome-wide association study data and novel computational methods to optimize polygenic risk scores (PRSs) for cardiometabolic disease and define their clinical utility in Arab people from Saudi Arabia
- The performance of currently used European-derived PRSs for coronary artery disease (CAD) and type 2 diabetes was substantially reduced in Arab people versus a matched cohort of Europeans from the UK Biobank
- Using public datasets, the PRSs were optimized for Arabs, and these scores performed significantly better than European-derived scores for predicting CAD, type 2 diabetes, and cardiomyopathy in Arab people
- Deciles of the optimized PRSs were clinically useful in stratifying Arab people for their risk of CAD and diabetes
- Increasing representation in genomic datasets should remain a priority, but existing resources could be helpful in cross-ancestry implementation of PRSs
Polygenic risk scores (PRSs) have been developed to identify individuals at risk of cardiometabolic disease, some cancers, and certain other diseases. However, these have been based mainly on European-ancestry populations in the U.S. and Europe. It's important to understand whether they are also useful for populations whose genetic ancestry, environmental factors, and disease epidemiology might differ.
Subscribe to the latest updates from Cardiovascular Advances in Motion
Arab people are massively underrepresented in genomic studies of cardiometabolic disease, even though some countries in the Arabian peninsula have the highest rates of diabetes and obesity worldwide. Along with colleagues in Saudia Arabia, researchers at Massachusetts General Hospital have optimized PRSs for cardiometabolic disease and validated their clinical utility in Arab people.
In Nature Communications, the research team describes how their procedure for optimizing PRSs, which leverages publicly available data and advances in computational methods, could be generalized to other populations.
The authors are Akl C. Fahed, MD, MPH, interventional cardiologist and physician–scientist in the Mass General Cardiovascular Research Center; Injeong Shim, formerly a visiting researcher in the Center; Xin Gao, MD, Hiroyuki Kuwahara, PhD, and NingNing Chen, of King Abdullah University of Science and Technology; Fowzan S. Alkuraya, MD, of King Faisal Specialist Hospital and Research Center; and colleagues.
Performance of European-derived PRSs
The study cohort comprised 5,399 individuals from all five regions of Saudi Arabia who had been referred for cardiology care at the King Faisal Specialist Hospital. The reference cohort was 1,017 individuals representative of the 28 tribes of indigenous Arabs in Saudi Arabia who were not known to have cardiometabolic disease.
In the disease cohort, the researchers analyzed 10 cardiometabolic traits/diseases, including coronary artery disease (CAD), type 2 diabetes, and cardiomyopathy, and continuous measures like lipid levels and blood pressure.
The reference cohort of Arab people was matched on age, sex, and each of the 10 diseases/traits to an equal number of individuals of European ancestry from the UK Biobank. The performance of the PRSs was:
- CAD—Odds ratio (OR) of disease per standard deviation (SD), 1.78 in Europeans vs. 1.41 in Arab people
- Type 2 diabetes—1.82 vs. 1.41
- Cardiometabolic traits—mean decrease in performance of 43% in Arab people
Performance of Arab-optimized PRSs
The researchers optimized the European-derived PRSs using public datasets. Specifically, for each disease/trait, they obtained the variant effect size from the largest and most diverse genome-wide association study and computed PRSs using five methods. Scores were then adjusted for principal components of ancestry.
In a training set of 2,700 patients in the disease cohort, the team identified the best-performing PRS for each disease/trait, and then tested it in the remaining 2,699 patients. The optimized scores performed markedly better for Arab people than the European-derived scores:
- CAD—OR of disease per SD, 1.41 for European score vs. 1.51 for optimized score
- Type 2 diabetes—1.41 vs. 1.83
- Cardiomyopathy—1.01 vs. 1.34
Similarly, Arab-optimized scores performed better than European scores for five of the seven cardiometabolic traits.
Clinical Utility of Arab-optimized PRSs
For each PRS, the researchers grouped the disease cohort into deciles of that score as defined using the reference population. In each decile, they identified the proportion of participants who had each disease of interest or had a value at or above the median of each trait.
The team noted marked stratification of disease risk by PRSs:
- CAD—OR, 3.94 for individuals in the top decile compared with those in the bottom decile
- Type 2 diabetes—OR, 8.19
There were also significant differences in continuous traits by PRS.
Advancing Equitable Implementation of PRSs
For some diseases/traits, the performance of Arab-optimized PRSs was on par with the previously published performance of PRSs in individuals of European ancestry. This suggests that even considering the limited genome-wide association data for Arab people and other non-European populations, the approach created for this study might facilitate cross-ancestry implementation of PRSs.
view original journal article Subscription may be required
Learn about the Corrigan Minehan Heart Center
Refer a patient to the Corrigan Minehan Heart Center