Assessing the Power of Exome Chips

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Genotyping chips for rare and low-frequency variants are increasingly popular but their utility is uncertain.
• These chips were designed using data from specific populations and diseases, potentially limiting their generalizability.
• Detecting associations with rare variants has lower statistical power compared to common variants.

Purpose of the Study:

• To empirically evaluate the power of exome chips for detecting genetic associations.
• To assess the impact of sample size, effect size, and variant scenarios on study power.
• To provide a simulation tool adaptable to various exome chip designs.

Main Methods:

• Developed a simulation program to model exome chip designs, specifically implementing the Illumina HumanExome BeadChip array.
• Assessed study power using simulated datasets across different effect sizes and causal variant scenarios.
• Applied two common statistical approaches for rare and low-frequency variants, including gene-based collapsing methods.

Main Results:

• Detecting modest effect sizes (0.5% < PAR > 1%) with 80% power requires 20,000–30,000 individuals under optimal conditions.
• Detecting small effect sizes (PAR <0.5%) requires 60,000–100,000 individuals, especially with non-causal variants present.
• Significant sample sizes are necessary for robust association studies using exome chips.

Conclusions:

• Tens of thousands of individuals are essential for detecting modest genetic effects with exome chips under ideal circumstances.
• Using rare variant chips on cohorts or diseases they were not designed for presents substantial challenges for variant or gene identification.
• The findings highlight the need for careful consideration of sample size and cohort matching when utilizing exome genotyping arrays.