SPC: a SPectral Component approach leveraging Identity-by-Descent graphs to address recent population structure in...

Area of Science:

• Statistical genetics
• Population genetics
• Bioinformatics

Background:

• Population structure is a critical confounder in genome-wide association studies (GWAS), potentially causing inflated test statistics and false positives.
• Traditional methods like principal components (PCs) struggle to capture fine-scale, non-linear population structures, limiting their effectiveness for rare variant analysis.

Purpose of the Study:

• To develop and validate a novel method, SPectral Components (SPCs), for accurately adjusting for fine-scale population structure in large genetic datasets.
• To compare the performance of SPCs against traditional PCs in accounting for population structure and its impact on genetic analyses.

Main Methods:

• Leveraging identity-by-descent (IBD) graphs to capture local, non-linear population structure.
• Transforming IBD graph information into continuous representations (SPCs) for integration into genetic analysis pipelines.
• Validating SPCs using simulated data and large-scale empirical data from the UK Biobank (N ≈ 420,000).

Main Results:

• SPCs explained over 90% of fine-scale population structure in simulations, significantly outperforming PCs ( < 50%).
• In the UK Biobank, SPCs reduced p-value inflation in GWAS by 12% more than PCs for an environmental phenotype.
• SPCs improved rare variant association analyses, reducing genomic inflation from 7.6 to 1.2, and provided more accurate heritability estimates.

Conclusions:

• SPCs offer a robust and scalable method for adjusting recent population structure, outperforming traditional PCs.
• SPCs enhance the accuracy of genome-wide association studies, particularly for rare variants and environmentally influenced traits.
• This novel approach provides a powerful alternative or complement to existing methods in large biobank studies.