Deshrinking ridge regression for genome-wide association studies.

Area of Science:

• Genetics and Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) are crucial for gene discovery, especially in the big data era with numerous traits.
• Existing methods like Efficient Mixed Model Association (EMMA) and Genome-wide Efficient Mixed Model Association (GEMMA) are widely used but computationally intensive.
• Ordinary Ridge Regression (ORR) shows similar test statistic patterns to EMMA but is limited by severe shrinkage effects.

Purpose of the Study:

• To develop a computationally efficient algorithm for GWAS.
• To adapt Ordinary Ridge Regression (ORR) for GWAS by addressing its shrinkage limitations.
• To create a method that is generalized across different model sizes and detects markers simultaneously.

Main Methods:

• Introduced a degree of freedom for each marker effect in ORR to deshrink estimated effects and standard errors.
• Developed the Deshrinking Ridge Regression (DRR) method, which adjusts ORR's Wald test to match EMMA's performance.
• Evaluated DRR against EMMA across small, medium, and large model sizes.

Main Results:

• DRR demonstrated superior generalization across all model sizes, unlike EMMA which is limited to medium and large models.
• DRR identifies all markers simultaneously, contrasting with the one-at-a-time scanning of other methods.
• DRR exhibits significantly simpler computational time complexity compared to EMMA and GEMMA, particularly when sample size is smaller than the number of genetic variants.

Conclusions:

• DRR provides a computationally efficient and generalized approach for GWAS.
• The method effectively overcomes the shrinkage limitations of ORR for genetic association analysis.
• DRR offers a promising alternative for large-scale genetic studies requiring high computational efficiency.