One score to rule them all: regularized ensemble polygenic risk prediction with GWAS summary statistics

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Ensemble learning enhances polygenic risk score (PRS) predictive power, commonly used in multi-ancestry PRS.
• Current ensemble methods require individual-level data, limiting application in underrepresented populations.
• Genomic data scarcity in non-European ancestries hinders PRS development.

Purpose of the Study:

• To develop a statistical framework for regularized ensemble PRS using only summary statistics.
• To enable PRS model combination without individual-level genetic data.
• To improve PRS prediction performance, especially in diverse populations.

Main Methods:

• Developed a novel statistical framework for regularized ensemble PRS construction.
• Utilized summary statistics from genome-wide association studies (GWAS) for model training.
• Combined a large number of candidate PRS models efficiently.

Main Results:

• Demonstrated robust and substantial improvement over existing PRS models.
• Achieved significant gains in both within-ancestry and cross-ancestry prediction.
• The proposed method shows superior performance compared to traditional PRS.

Conclusions:

• The regularized ensemble PRS framework offers a powerful, data-efficient approach.
• This method overcomes limitations of individual-level data requirements.
• It provides a universal, continuously improvable solution for future PRS applications.