Transfer learning for genomic prediction in underrepresented populations
Google Research Blog
Read full postResearchers studied how polygenic risk scores (PRSs) derived from large European genome-wide association studies (GWASs) perform in predicting disease risk in non-European populations, focusing on transfer learning to improve accuracy. Using UK Biobank data and Biobank Japan, they evaluated PRS accuracy across eight clinical traits and varying sample sizes, aiming to optimize predictive models for underrepresented groups.





