Local True Discovery Rate Weighted Polygenic Scores Using GWAS Summary Data.

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Polygenic scores (PS) are vital for assessing individual genetic disease risk using genome-wide data.
• Standard PS construction relies on SNP weights derived from regression coefficients, often using arbitrary p-value thresholds.
• Current methods face challenges like multiple comparisons and overfitting due to threshold selection.

Purpose of the Study:

• To propose an alternative weighting scheme for polygenic scores.
• To improve the accuracy of genetic risk prediction.
• To address limitations associated with p-value thresholding in PS construction.

Main Methods:

• Developed a novel weighting scheme for PS utilizing local true discovery rates (LTDR).
• Estimated LTDR using maximum likelihood and kernel density estimation from summary statistics.
• Validated the proposed method through simulation studies with real and artificial genetic data.

Main Results:

• The proposed LTDR-based weighting scheme demonstrated predictive performance comparable to standard PS.
• The new method achieved high accuracy even without knowing the optimal p-value threshold in advance.
• Simulation results showed the robustness of the LTDR approach in genetic prediction.

Conclusions:

• The novel LTDR-based weighting scheme offers a competitive alternative for constructing polygenic scores.
• This approach mitigates issues related to p-value threshold selection in PS.
• The method holds promise for more reliable genetic risk prediction in various diseases.