SNP set association analysis for genome-wide association studies

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Genome-wide association studies (GWAS) identify genetic variants linked to diseases using millions of single nucleotide polymorphisms (SNPs).
• Single locus association studies face challenges with low power due to multiple comparison corrections.
• Grouping SNPs into sets based on genomic features can improve power by testing joint effects.

Purpose of the Study:

• To compare the performance of Principal Component Analysis (PCA), Supervised Principal Component Analysis (SPCA), Kernel Principal Component Analysis (KPCA), and Sliced Inverse Regression (SIR).
• To evaluate these methods' ability to control Type I error rates and enhance power in identifying disease-associated genetic variants.

Main Methods:

• Simulated SNP sets were generated under models with 0, 1, and ≥ 2 causal SNPs.
• Evaluated performance metrics including Type I error control and statistical power.
• Applied four distinct dimensionality reduction and association testing methods: PCA, SPCA, KPCA, and SIR.

Main Results:

• All tested methods effectively controlled Type I error rates at the nominal significance level.
• SPCA demonstrated consistently higher power compared to PCA, KPCA, and SIR across various settings.
• Performance variations were observed based on linkage disequilibrium structures and minor allele frequencies in simulated data.

Conclusions:

• SPCA is a more powerful approach for SNP set-based association testing in GWAS.
• The findings support SPCA's utility for identifying complex genetic architectures of diseases.
• The methods were validated on a real-world GWAS dataset for non-small cell lung cancer (NSCLC) in the Han Chinese population.