Controlling for human population stratification in rare variant association studies.

Area of Science:

• Population genetics
• Statistical genetics
• Genomic association studies

Background:

• Population stratification is a significant confounder in genetic association studies, particularly for rare variants.
• Existing correction methods like principal components (PCs) and linear mixed models (LMMs) can yield conflicting results, especially under complex stratification scenarios.
• Previous evaluations often used limited stratification types and large sample sizes, not fully addressing analyses of rare disorders.

Purpose of the Study:

• To investigate the performance of different population stratification correction methods in genetic association studies.
• To evaluate these methods across various stratification scenarios (within- and between-continent) and sample sizes, including those with few cases (down to 50).
• To identify robust methods for controlling type-I errors and enhancing statistical power in rare disease genetic analyses.

Main Methods:

• A large-scale simulation study using real exome data was conducted.
• Multiple correction methods, including principal components (PCs), linear mixed models (LMMs), and a novel local permutation method (LocPerm), were evaluated.
• Scenarios included diverse sample sizes, varying numbers of cases and controls, and different population structures (worldwide, continental, within-continent).

Main Results:

• Principal components (PCs) showed inflated type-I errors with few controls (≤100) in samples of 50 cases.
• Linear mixed models (LMMs) exhibited inflated type-I errors with many controls (≥1000) in samples of 50 cases.
• The novel local permutation method (LocPerm) consistently maintained correct type-I error rates across all tested scenarios.
• Statistical power was comparable across methods, highlighting the primacy of accurate type-I error control.
• Adding external controls to analyses with few cases significantly increased power when appropriate stratification correction was applied.

Conclusions:

• The local permutation (LocPerm) method offers a reliable approach for controlling type-I errors in genetic association studies with population stratification.
• Standard methods like PCs and LMMs may be unreliable under specific sample size and stratification conditions, particularly for rare disease studies.
• Leveraging large external control panels can enhance the power of genetic association studies for rare disorders, provided robust stratification correction is employed.