Principal Component Analysis Reduces Collider Bias in Polygenic Score Effect Size Estimation.

Area of Science:

• Genetics
• Statistical Genetics
• Epidemiology

Background:

• Collider bias can distort genetic association findings.
• Polygenic risk scores (PRS) are widely used but susceptible to bias.
• Accurate estimation of genetic effects is crucial for understanding disease etiology.

Purpose of the Study:

• To evaluate principal component analysis (PCA) of measured confounders as a method to mitigate collider bias in polygenic association models.
• To assess the performance of PCA in reducing collider bias using simulations and real-world genetic data.
• To investigate the impact of PCA on polygenic score effect size estimation.

Main Methods:

• Applied principal component analysis (PCA) to measured confounders in polygenic association models.
• Conducted simulations to test the efficacy of PCA under various assumptions.
• Utilized data from the Collaborative Study of the Genetics of Alcoholism (COGA) with a polygenic score for alcohol problems and alcohol use disorder as the phenotype.
• Examined tobacco use and educational attainment as collider variables.

Main Results:

• Simulation results indicate that assumptions about confounder correlation structure and availability are complementary.
• PCA effectively reduced collider bias in the COGA sample when tobacco use was the collider variable.
• The method demonstrated potential for improving PRS effect size estimation by minimizing collider bias.

Conclusions:

• Principal component analysis (PCA) of measured confounders is a viable strategy to reduce collider bias in genetic studies.
• This approach offers an efficient way to leverage existing data resources for more accurate genetic effect estimation.
• The findings have implications for improving the reliability of polygenic risk scores in diverse populations.