Characterizing the Uncertainty, Misclassification and Inconsistency of Polygenic Prediction

Polygenic risk scores (PRSs) hold promise for precision medicine, yet their clinical translation is hindered by substantial uncertainty in individual risk estimates and often limited agreement in risk stratification across multiple PRSs for the same disease. We develop a unified inferential framework to calibrate PRS point estimates and uncertainties for both quantitative traits and binary phenotypes, and to characterize how PRS accuracy, uncertainty, pairwise correlation jointly determine misclassification and classification inconsistency. We show, both theoretically and empirically, that individual- and population-level misclassification and inconsistency rates are highly predictable in independent datasets. We further evaluate PRS integration and uncertainty-aware probabilistic thresholding strategies that reduce misclassification and improve concordance in risk stratification. Together, these results demonstrate that instability in PRS-based classification is a predictable statistical consequence of uncertainty and establish a principled foundation for incorporating uncertainty into PRS-based risk interpretation, communication, and clinical decision-making.