A Principled Framework for Using Correlated Traits to Improve Risk Prediction
Although many complex phenotypes and diseases are influenced by shared genetic and environmental factors, risk prediction methods typically rely on genetic information from a single trait, leaving a rich source of predictive information largely unexploited. Phenotypic correlations can potentially be used to improve the accuracy of polygenic scores (PGS), but the conditions under which correlated…
In a groundbreaking study, researchers have unveiled a principled framework for utilizing correlated traits to enhance risk prediction. Many complex phenotypes and diseases arise from shared genetic and environmental factors, yet traditional risk prediction methods usually focus on a single trait, leaving untapped potential in predictive insights. By harnessing the power of phenotypic correlations, a new approach has emerged that can significantly boost the accuracy of polygenic scores (PGS).
To better understand when and how correlated traits can meaningfully improve prediction, the researchers developed a comprehensive theoretical and simulation framework. This framework quantifies the extent to which helper traits—additional traits that provide predictive information—enhance the accuracy of PGS models. The framework also identifies the critical factors that determine the magnitude of these gains, offering valuable guidance for selecting the most informative helper traits.
The findings reveal that helper traits can substantially improve predictive accuracy, with the extent of improvement dependent on various factors. These factors include the baseline performance of the model, genetic and environmental correlations between the target and helper traits, and the heritability of both the target and helper traits.
Surprisingly, even when the target trait has a weak heritability, helper traits can still substantially boost PGS accuracy. This is because low-heritability traits may capture non-redundant environmental factors that are shared between the target and helper traits.
To validate the efficacy of this approach, the researchers empirically evaluated the use of helper traits by developing PGS models to predict type 2 diabetes using data from the UK Biobank. The results demonstrated that incorporating helper traits led to a substantial improvement in predictive accuracy compared to relying on a single-trait PGS alone.
Specifically, the AUC-ROC (area under the receiver operating characteristic curve) improved from 0.677 for the single-trait PGS to 0.907 with the inclusion of helper traits. Notably, the accuracy achieved using helper traits was comparable to that of models utilizing HbA1c, which is currently considered the clinical gold standard biomarker for type 2 diabetes.
These findings establish a robust theoretical and practical framework for leveraging correlated traits to enhance polygenic prediction. The study provides clear guidance on selecting informative helper traits based on empirically measurable quantities, enabling researchers to maximize the benefits of shared genetic and environmental architectures.
Furthermore, the researchers have developed an interactive web application that allows users to estimate the expected gain in accuracy from candidate helper traits, based on readily available data.
This groundbreaking research paves the way for a more comprehensive and accurate approach to risk prediction, unlocking the potential of correlated traits to improve outcomes in various complex diseases and phenotypes. By harnessing the power of shared genetic and environmental factors, this novel framework promises to revolutionize our understanding and prediction of complex traits, ultimately leading to better preventive strategies and personalized treatment plans.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
