Quantification of population structure using correlated SNPs by shrinkage principal components

Area of Science:

• Genetics
• Population Genetics
• Statistical Genetics

Background:

• Genome-wide association studies (GWAS) are crucial for mapping complex traits.
• Population stratification is a major challenge in association mapping, often controlled using Principal Component Analysis (PCA).
• Standard PCA assumes markers are in linkage equilibrium, which is often not true.

Purpose of the Study:

• To investigate the impact of linkage disequilibrium (LD) on PCA for population stratification control.
• To develop a modified PCA method that accounts for marker correlations due to LD.

Main Methods:

• Examined the influence of LD on PCA performance in GWAS datasets.
• Proposed a modified PCA approach to automatically adjust for marker correlations.

Main Results:

• Demonstrated that LD patterns can create artifactual subpopulations in standard PCA.
• Showed the proposed method effectively removes LD-induced artifacts.
• Successfully recovered true population structure obscured by LD in standard PCA.

Conclusions:

• High LD among SNPs significantly impacts PCA, masking true population substructure.
• Shrinkage PCA is robust to LD patterns and applicable to all markers.
• The proposed method offers a simpler alternative to existing LD-adjusted PCA techniques.