Multi-Omics Integration Improves Polygenic Risk Prediction for Lipid Traits: A Multi-Ancestry Study in UK Biobank

Area of Science:

• Genomics
• Proteomics
• Metabolomics
• Cardiovascular Disease Research

Background:

• Polygenic risk scores (PRS) are useful for disease prediction but limited by complex gene-environment interactions.
• Blood lipid levels are heritable cardiovascular disease risk factors, but multi-omics integration's role in prediction is unclear.
• Enhancing lipid trait prediction beyond PRS alone is crucial for cardiovascular health.

Purpose of the Study:

• To investigate if multi-omics integration enhances lipid trait prediction compared to PRS alone.
• To develop and evaluate composite multi-omics risk scores (MoRS) for lipid traits.
• To assess predictive accuracy across diverse ancestries.

Main Methods:

• Constructed single-omics scores using genetically predicted gene expression, plasma proteins, and metabolites, weighted by LASSO regression.
• Developed composite multi-omics risk scores (MoRS) using stepwise regression (step-MoRS) and direct LASSO modeling (Lasso-MoRS).
• Evaluated MoRS performance across European, South Asian, and African ancestries in the UK Biobank.

Main Results:

• Multi-omics risk scores (MoRS) consistently outperformed PRS alone in predicting four lipid traits across diverse populations.
• Lasso-MoRS identified key biomarkers that complement genomic data for improved prediction.
• Enhanced predictive accuracy was observed across European, South Asian, and African ancestries.

Conclusions:

• Integrating multi-omics biomarkers with genomic data significantly improves lipid trait prediction across ancestries.
• This multi-omics approach provides biological insights into lipid metabolism regulation.
• Multi-omics integration offers a promising strategy for refining cardiovascular disease risk prediction.