Assessing statistical significance in multivariable genome wide association analysis

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome Wide Association Studies (GWAS) typically analyze single nucleotide polymorphisms (SNPs) individually.
• This marginal analysis approach has limited power due to the low predictive ability of single SNPs and stringent multiple testing correction.
• This can lead to a decreased ability to detect true genetic associations with diseases.

Purpose of the Study:

• To develop and validate a novel statistical procedure for analyzing high-dimensional GWAS data.
• To improve the power and accuracy of identifying significant SNPs and SNP groups associated with complex diseases.
• To address the limitations of traditional single-SNP analysis in GWAS.

Main Methods:

• Proposed a multiple generalized linear model for simultaneous analysis of all SNPs.
• Developed a method to compute P-values for single SNPs or SNP groups while controlling the family-wise error rate (FWER).
• Implemented the procedure for extremely high-dimensional datasets and validated it using WTCCC data.

Main Results:

• The proposed method effectively analyzes all SNPs in a multiple generalized linear model, suitable for high-dimensional data.
• It provides P-values for SNP significance, controlling for all other SNPs and the FWER, thus avoiding spurious correlations.
• The method identified significant SNPs not found in the original WTCCC study but replicated in independent studies, demonstrating its enhanced discovery power.

Conclusions:

• The developed method offers a powerful and robust approach for GWAS analysis, outperforming traditional marginal methods.
• It accurately identifies SNPs and SNP groups with genuine associations to phenotypes by considering joint effects.
• The open-source package hierGWAS ensures reproducibility and facilitates the application of this advanced GWAS analysis technique.