Many Phenotypes Without Many False Discoveries: Error Controlling Strategies for Multitrait Association Studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Investigating the genetic basis of multiple phenotypes requires testing numerous hypotheses, linking traits to genetic variants.
• Controlling for multiple testing is essential for reliable findings, with the false discovery rate (FDR) commonly used.
• Current FDR methods can overstate discoveries when applied to large-scale genetic association studies of multiple traits.

Purpose of the Study:

• To address the limitations of standard FDR procedures in controlling false discoveries in multi-phenotype genetic association studies.
• To propose a novel hierarchical testing procedure for more accurate control of error rates.
• To provide a robust framework for identifying genetic variants with functional effects on multiple traits.

Main Methods:

• Developed a hierarchical testing procedure designed for multi-phenotype genetic association analyses.
• Conducted simulation studies to compare the proposed method against existing FDR controlling procedures.
• Evaluated various error rates and statistical power metrics in simulated genetic association studies.

Main Results:

• Standard FDR control methods were shown to inadequately control the rate of false variant discoveries and the proportion of falsely associated phenotypes.
• The proposed hierarchical testing procedure effectively controls both the rate of false variant discoveries and the average proportion of falsely associated phenotypes.
• Simulations demonstrated superior performance of the hierarchical method in controlling error rates and maintaining power.

Conclusions:

• A new hierarchical testing approach offers improved control over false discoveries in multi-phenotype genetic studies.
• This method enhances the reliability of identifying genetic variants associated with complex traits.
• Application to Arabidopsis thaliana flowering time identified novel genetic variants impacting phenotypes.