Polygenic Prediction of Type 2 Diabetes in Africa

Abstract

Objective: Polygenic prediction of type 2 diabetes (T2D) in continental Africans is adversely affected by the limited number of genome-wide association studies (GWAS) of T2D from Africa and the poor transferability of European-derived polygenic risk scores (PRSs) in diverse ethnicities. We set out to evaluate if African American, European, or multiethnic-derived PRSs would improve polygenic prediction in continental Africans.

Research design and methods: Using the PRSice software, ethnic-specific PRSs were computed with weights from the T2D GWAS multiancestry meta-analysis of 228,499 case and 1,178,783 control subjects. The South African Zulu study (n = 1,602 case and 981 control subjects) was used as the target data set. Validation and assessment of the best predictive PRS association with age at diagnosis were conducted in the Africa America Diabetes Mellitus (AADM) study (n = 2,148 case and 2,161 control subjects).

Results: The discriminatory ability of the African American and multiethnic PRSs was similar. However, the African American-derived PRS was more transferable in all the countries represented in the AADM cohort and predictive of T2D in the country combined analysis compared with the European and multiethnic-derived scores. Notably, participants in the 10th decile of this PRS had a 3.63-fold greater risk (odds ratio 3.63; 95% CI 2.19-4.03; P = 2.79 × 10-17) per risk allele of developing diabetes and were diagnosed 2.6 years earlier than those in the first decile.

Conclusions: African American-derived PRS enhances polygenic prediction of T2D in continental Africans. Improved representation of non-European populations (including Africans) in GWAS promises to provide better tools for precision medicine interventions in T2D.

Figure 1

A: Shape plot for the difference in odds ratio for T2D (adjusted for age, sex, BMI, and five principal components) in reference to the first decile for the African American PRS in the AADM study. B: Bar plots showing the transferability of the PRS in African countries represented in the AADM study.

Figure 2

A: Receiver operating curves for the African American–derived PRS and conventional risk factors for the prediction of T2D in the AADM study. Full model refers to age, sex, BMI, African American PRS, and 5PCs. B: Shape plot for the difference of age at diagnosis for T2D in the AADM study for the African American–derived PRS. 5PCs, five principal components.