Accurate error control in high-dimensional association testing using conditional false discovery rates.

Area of Science:

• Biostatistics
• Genomics
• Biomedical Sciences

Background:

• High-dimensional hypothesis testing is crucial in biomedical research.
• Informative covariates can enhance statistical power.
• Conditional false discovery rate (cFDR) is a common approach, but existing methods have limitations in type-1 error control and can be over-conservative.

Purpose of the Study:

• To develop a new method for type-1 error rate control in cFDR analysis.
• To improve the power of cFDR analysis using informative covariates.
• To enhance the applicability of cFDR analysis in complex genomic studies.

Main Methods:

• Proposed a novel type-1 error rate control method based on estimated cFDR mappings.
• Developed an adjustment to the existing cFDR estimator to further improve power.
• Validated the method through simulations and applied it to transcriptome-wide association studies (TWAS).

Main Results:

• The new method demonstrated more than double the potential improvement in power compared to existing methods.
• The proposed approach offers better type-1 error rate control.
• The method was successfully applied to TWAS, showing substantial improvements in power and applicability.

Conclusions:

• The developed method offers significant advancements in high-dimensional hypothesis testing.
• It provides improved power and robust type-1 error rate control for cFDR analysis.
• The iterative application of the method allows for the successive use of multiple covariates, increasing its utility in complex analyses.