25 November a 15 minute ‘Focus Talks’ between 11:00 – 11:45am. Polygenic risk scores (PRS) utilize the results from GWAS to quantify disease risks. A major challenge for translational applications of PRS is that the majority of GWAS come from cohorts of European ancestry, and the predictive power for individuals of non-European or mixed ancestry is unclear. In this talk, we will review geographic differences of PRS risk predictions for coronary artery disease, rheumatoid arthritis, schizophrenia, and further disorders from the literature. We will then explain by simulated data how differences in allele frequencies, linkage disequilibrium, and phenotypic variance across ancestries can affect transferability. We will argue how better fine-mapping resolution, could improve identification of single causal variants, that would make PRS models more robust. However, we will also demonstrate how differences in effect sizes across populations, would reduce the predictive value even if all causal sites are known. We conclude that for PRS in clinical practice, models should ideally be based on the patient population. Prof. Peter Krawitz, Head of Institute for Genomic Statistics and Bioinformatics (W3), University Bonn, Germany |