Protein prediction for trait mapping in diverse populations

Area of Science:

• Genomics and Proteomics
• Precision Medicine
• Complex Trait Genetics

Background:

• Genetically regulated gene expression aids in understanding complex traits.
• High-throughput technology enables proteome interrogation for similar insights.
• The Trans-omics for Precision Medicine (TOPMed) Multi-omics pilot study provides valuable data.

Purpose of the Study:

• To optimize genetic predictors of the plasma proteome for proteome-wide association studies (PWAS) in diverse populations.
• To assess the transferability of predictive models across different ancestries.
• To identify protein-trait associations for complex traits using PWAS.

Main Methods:

• Utilized TOPMed Multi-Ethnic Study of Atherosclerosis (MESA) data to build predictive models for 1,305 proteins.
• Compared elastic net regression models with and without fine-mapping using posterior inclusion probabilities.
• Applied S-PrediXcan to GWAS summary statistics from the Population Architecture using Genomics and Epidemiology (PAGE) study for PWAS.

Main Results:

• Fine-mapping generated more significant protein prediction models, particularly in African ancestries, increasing discovery potential.
• Models trained in TOPMed MESA showed improved cross-ancestry prediction in the INTERVAL study when fine-mapping was used.
• PWAS identified protein-trait associations that colocalized and replicated in independent GWAS, especially when training populations matched PAGE ancestries.

Conclusions:

• Optimized genetic predictors for plasma proteome enable PWAS in diverse populations.
• Fine-mapping enhances protein prediction models, aiding in the discovery of genetic associations with complex traits.
• Publicly available predictive models facilitate proteome mapping research.