Quantifying Uncertainty in Polygenic Risk Scores Using Conformalized Quantile Regression

Polygenic risk scores (PRS) are widely used in post-GWAS analyses to predict complex traits across humans, animals, and plants. While significant progress has been made in developing new PRS methods, much less attention has been given to quantifying the uncertainty associated with these predictions. In this work, we explore a method for uncertainty quantification based on quantile regression. When paired with conformal prediction, this approach enables the construction of prediction intervals with guaranteed coverage, offering lower and upper bounds within which the phenotype is likely to fall with high probability. We apply this framework to data from the UK Biobank and the ProgeNIA/SardiNIA studies, showing that the resulting prediction intervals: (1) exhibit reduced uncertainty compared to standard methods, (2) maintain valid coverage under minimal model assumptions, and (3) provide more realistic estimates by allowing for asymmetry and individual-specific interval lengths. Overall, we present a novel framework for uncertainty quantification in PRS analyses and highlight the importance of incorporating uncertainty into predictive modeling.