Polygenic risk modeling with latent trait-related genetic components

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Polygenic risk scores (PRS) are valuable for predicting complex disease outcomes.
• Standard PRS often lack the granularity to explain within-trait diversity, such as disease subtypes or underlying biological pathways.

Purpose of the Study:

• Introduce a novel latent factor model, the Decompositions of Genetic Associations polygenic risk score (dPRS).
• Enhance the interpretability of genetic risk for complex traits by decomposing it into meaningful components.

Main Methods:

• Computed Decompositions of Genetic Associations (DeGAs) using genetic associations across 977 traits.
• Developed and applied the dPRS model to UK Biobank data (337,151 individuals) and a replication cohort (25,486 individuals).
• Decomposed individual genetic risk for traits like body mass index (BMI) and myocardial infarction.

Main Results:

• The dPRS demonstrated comparable predictive performance to standard PRS.
• dPRS provides enhanced interpretability, decomposing BMI genetic risk into components related to fat-free mass, fat mass, and physical activity.
• Identified distinct patterns of genetic risk contribution for BMI, including individuals with contributions from single components.

Conclusions:

• The dPRS method offers a powerful tool for fine-scale interpretation of genetic risk factors in complex traits.
• This approach advances our understanding of the biological heterogeneity underlying common diseases.
• dPRS facilitates a more nuanced view of how genetic predispositions manifest in individuals.