Leveraging both individual-level genetic data and GWAS summary statistics increases polygenic prediction

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Polygenic risk scores (PRSs) predict complex diseases, with accuracy increasing with training data size.
• Traditionally, PRSs use summary statistics from genome-wide association studies (GWASs) meta-analyses.
• Large individual-level datasets (e.g., UK Biobank) are increasingly available.

Purpose of the Study:

• To investigate optimal methods for combining summary statistics and individual-level data for polygenic prediction.
• To compare the performance of meta-GWAS with alternative data-combining strategies.

Main Methods:

• Simulations were conducted.
• Real data from 12 case-control and quantitative traits (iPSYCH and UK Biobank) were used.
• Meta-genome-wide association study (meta-GWAS) was compared against stacked clumping and thresholding (SCT) and meta-polygenic risk score (meta-PRS).

Main Results:

• Meta-GWAS does not always yield the most accurate PRS.
• The meta-PRS approach, a linear combination of PRSs, was evaluated.
• Meta-PRS demonstrated comparable or superior accuracy to meta-GWAS when large individual-level data were available.

Conclusions:

• Combining summary statistics and individual-level data can enhance PRS accuracy.
• Meta-PRS presents a simple and effective alternative to meta-GWAS for polygenic prediction.
• The findings suggest a new standard for leveraging diverse genetic data sources.