FairPRS: adjusting for admixed populations in polygenic risk scores using invariant risk minimization

• DOI: 10.1142/9789811270611_0019
• Publication Date: 2022-11-24T08:58:01Z
• Authors: Diego Machado Reyes; Aritra Bose; Ehud Karavani; Laxmi Parida
• Keywords: Polygenic risk score; Minification

Polygenic risk scores (PRS) are increasingly used to estimate the personal of a trait based on genetics. However, most genomic cohorts European populations, with strong under-representation non-European groups. Given that PRS poorly transport across racial groups, this has potential exacerbate health disparities if in clinical care. Hence there is need generate perform comparably ethnic Borrowing from recent advancements domain adaption field machine learning, we propose FairPRS - an Invariant Risk Minimization (IRM) approach for estimating fair or debiasing pre-computed PRS. We test our method both diverse set synthetic data and real UK Biobank. show can create ancestry-invariant distributions racially unbiased largely improve phenotype prediction. hope will contribute fairer characterization patients by genetics rather than race.