Existing genome-wide association studies have been predominantly conducted in individuals of European descent. In turn, this has limited the transferability of polygenic risk scores (PRS) in non-European populations. Researchers have now presented a novel PRS method – PRS-CSx – which improves cross-population polygenic prediction.
Polygenic risk scores
Most human traits and diseases are complex and influenced by hundreds or thousands of genetic variants, with each explaining a small proportion of phenotypic variation. PRS aggregates the effects of genetic variants across the genome to measure the overall genetic liability of a trait or disease. PRS have shown particular promise in predicting individualised disease risk and trajectories, stratifying patient groups, informing preventive, diagnostic and therapeutic strategies and also improving health outcomes.
Despite the potential for clinical translation, recent studies have shown that PRS have decreased cross-population prediction accuracy. Due to existing GWAS analyses predominantly involving individuals of European descent, poor transferability of PRS across populations has impeded its clinical implementation and raised health disparity concerns. As a result, there is a urgent need to improve the accuracy of cross-population polygenic prediction in order to maximise the clinical potential of PRS and ensure equity of precision medicine.
In a study, published as preprint in medRxiv, researchers have presented a novel PRS method – PRS-CSx – an extension of PRS-CS (Bayesian polygenic prediction method). This method jointly models GWAS summary statistics from multiple populations. It assumes that causal variants are largely shared across populations while allowing for their effect sizes to vary. Moreover, PRS-CSx also inherits the advantages of continuous shrinkage priors in polygenic modelling and prediction from PRS-CS. This enables accurate modelling of population-specific LD patterns and computationally efficient posterior inference.
The team compared the predictive performance of PRS-CSx with existing PRS construction methods across traits with a wide range of genetic architectures, cross-population genetic correlations and discovery GWAS samples sizes via simulations. They further applied this method to predict quantitative traits from various biobanks. They found that PRS-CSx outperforms existing methods across traits with a wide range of genetic architectures and cross-population genetic correlations in simulation. Additionally, the method substantially improved the prediction of quantitative traits and schizophrenia risk in non-European populations.
PRS-CSx can greatly improve cross-population polygenic prediction. Nonetheless, the gap in prediction accuracy between European and non-European populations remains notable. The team believe that the rapid expansion of genomic resources for non-European populations combined with advanced analytic methods will accelerate the translation of PRS in clinical settings and maximise its healthcare potential.
Image credit: By kjpargeter – www.freepik.com