Winner's Curse Correction and Variable Thresholding Improve Performance of Polygenic Risk Modeling Based on...

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Genome-wide association studies (GWAS) provide summary-level data for polygenic risk score (PRS) development.
• Standard PRS methods have limitations in accurately predicting genetic risk for complex diseases.
• Heritability analyses suggest potential for PRS improvement using GWAS data.

Purpose of the Study:

• To enhance the performance of polygenic risk scores (PRS).
• To introduce novel methods for improving genetic risk prediction accuracy.
• To refine SNP weighting and selection strategies in PRS construction.

Main Methods:

• Implemented threshold-dependent winner's-curse adjustments for association coefficients.
• Incorporated variable thresholds for SNP selection based on functional/annotation knowledge.
• Applied proposed methods to GWAS summary-level data across 14 complex diseases.

Main Results:

• Winner's curse correction uniformly improved PRS model performance across all tested diseases.
• Incorporating functional SNPs enhanced prediction for specific diseases.
• Combined methods achieved a 25-50% increase in prediction R2 for 5 out of 14 diseases.
• Type 2 diabetes GWAS showed improved prediction R2 from 2.29% to 3.53% with proposed methods.

Conclusions:

• Modified PRS methods offer significant gains in genetic risk prediction efficiency.
• Winner's curse correction is a universally beneficial adjustment for PRS.
• Functional SNP incorporation shows disease-specific benefits, warranting careful application.
• Understanding linkage disequilibrium is crucial for optimizing functional SNP utilization in PRS.