Experimental designs for robust detection of effects in genome-wide case-control studies

Area of Science:

• Genetics
• Statistical Genetics
• Genomic Epidemiology

Background:

• Genome-wide association studies (GWAS) aim to identify genomic loci associated with diseases by scanning hundreds of thousands of markers in thousands of individuals.
• A significant portion of common disease heritability remains unexplained, often referred to as the "dark matter of the genome."
• Robust detection of associations requires high statistical power and strong evidence (e.g., Bayes factor > 10^6) to overcome low prior probabilities.

Purpose of the Study:

• To establish experimental design criteria for robustly detecting associations between genomic loci and phenotypes in GWAS.
• To provide power calculations for determining necessary sample sizes to detect genetic effects with high confidence.

Main Methods:

• Developed power calculations for detecting effects of biallelic markers in linkage disequilibrium with causal loci.
• Considered additive, dominant, and recessive genetic models.
• Implemented calculations in R, to be available in the ldDesign package.

Main Results:

• Significantly stronger statistical evidence and larger sample sizes are required for robust association detection than traditional methods suggest.
• Many previously reported putative genetic effects may not be robustly detected.
• Potentially large, low-frequency genetic effects may remain undetected, contributing to the unexplained heritability.

Conclusions:

• Current GWAS methodologies may be underpowered to detect many true genetic associations, particularly for complex diseases.
• The stringent evidence and sample size requirements highlighted may explain the persistent "dark matter" in human genetics.
• The developed power calculations and R package will aid researchers in designing more effective GWAS experiments.