Polygenic risk prediction based on singular value decomposition with applications to alcohol use disorder

Area of Science:

• Genetics
• Biostatistics
• Psychiatry

Background:

• Polygenic risk scores (PRS) offer a way to assess genetic susceptibility to complex diseases like alcohol use disorder.
• Conventional PRS methods can suffer from low predictive power due to over-adjustment of confounding factors in discovery samples.

Purpose of the Study:

• To develop and evaluate a novel PRS construction method that overcomes limitations of conventional approaches.
• To enhance the prediction accuracy of genetic risk for alcohol use disorder.

Main Methods:

• Proposed a new PRS method involving principal component approximation in discovery data and projection onto target data.
• Utilized the Study of Addiction: Genetics and Environment (SAGE) for discovery and the National Longitudinal Study of Adolescent to Adult Health (Add Health) for target data.
• Compared the performance of the proposed PRS method against conventional methods.

Main Results:

• The proposed PRS method demonstrated superior prediction power compared to conventional methods.
• The new method showed improved performance in handling participants from diverse ancestry backgrounds.
• Provided recommendations for optimizing linkage disequilibrium (LD) and p-value thresholds.

Conclusions:

• The novel PRS method offers a more robust and accurate approach for summarizing genetic risk.
• This method has the potential to improve the application of PRS in diverse populations.
• Practical guidelines for PRS parameter selection were established.