Computationally efficient permutation-based confidence interval estimation for tail-area FDR.

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Genomic studies face challenges with parametric assumptions and numerous tests.
• Traditional False Discovery Rate (FDR) methods combined with permutation testing are computationally intensive.
• There is a need for efficient and reliable FDR estimation in high-throughput genomic analyses.

Purpose of the Study:

• To develop a computationally efficient permutation-based approach for FDR estimation in genomic studies.
• To provide tractable estimators for the proportion of true null hypotheses and the variance of the log of tail-area FDR.
• To introduce a confidence interval (CI) estimator for FDR that accounts for permutation number and test dependencies.

Main Methods:

• A novel permutation-based approach incorporating estimators for true null proportion and FDR variance.
• A CI estimator utilizing binomial distribution and an overdispersion parameter for positive test counts.
• Validation of the method's performance and reliability with as few as 10 permutations.

Main Results:

• The proposed method demonstrates favorable performance compared to existing approaches.
• Reliable FDR estimates are achievable with a minimal number of permutations (e.g., 10).
• Application to sleep and gene expression data identified 11 transcripts associated with REM sleep (FDR = 0.15), including genes involved in wnt and interferon signaling, which might be missed with stricter thresholds.

Conclusions:

• The developed method offers a computationally efficient and reliable alternative for FDR estimation in genomics.
• The CI provides flexibility in setting significance thresholds, enabling data-driven approaches for identifying weak but potentially significant effects.
• This approach is particularly valuable for studies of complex diseases where effect sizes are often small.