Using Alternative Definitions of Controls to Increase Statistical Power in GWAS.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) face challenges with statistical power due to small effect sizes of single nucleotide polymorphisms (SNPs) and stringent multiple testing criteria.
• Increasing sample size is the conventional method to boost GWAS power, but this can be resource-intensive.

Purpose of the Study:

• To propose and evaluate an alternative strategy for enhancing GWAS statistical power by redefining case-control outcomes into an ordinal classification.
• To compare the power of an ordinal model against standard case-control and case-asymptomatic control designs.

Main Methods:

• A simulation study was conducted to assess statistical power under varying effect sizes, minor allele frequencies, population prevalences, and subthreshold group sizes.
• The simulation included three scenarios: standard case-control, ordinal (case-subthreshold-asymptomatic), and case-asymptomatic control analyses.
• A real-world dataset analyzing major depression from the UK Biobank was used to validate simulation findings.

Main Results:

• The ordinal model consistently demonstrated superior statistical power compared to the standard case-control model.
• The case-asymptomatic control model's power varied, depending on population prevalence and the size of the subthreshold group.
• Analysis of major depression data supported the simulation results, indicating an approximate 10% increase in GWAS power with the ordinal model.

Conclusions:

• Redefining control groups into an ordinal classification (case-subthreshold-asymptomatic) significantly enhances statistical power in GWAS.
• The ordinal approach offers a cost-effective alternative to increasing sample size for improving GWAS discovery.
• This method holds promise for identifying genetic associations with complex traits and diseases.