A statistical method for the conservative adjustment of false discovery rate (q-value)

Area of Science:

• Genomics
• Statistical Bioinformatics
• Computational Biology

Background:

• q-value is a standard method for estimating false discovery rate (FDR) in genome-wide expression data analysis.
• Existing q-value methods may underestimate FDR, leading to false discoveries, particularly when permutation procedures are needed for p-value calculation.
• This underestimation issue is not well-addressed in current literature.

Purpose of the Study:

• To propose a statistical method for the conservative adjustment of q-value.
• To address the underestimation of FDR in genomic data analysis, especially when permutation tests are employed.
• To ensure more reliable identification of significant findings in high-throughput biological data.

Main Methods:

• Development of a novel statistical adjustment method for q-value calculation.
• Incorporation of permutation procedures for p-value calculation into the adjustment method.
• Validation using simulation data and experimental microarray and sequencing data.

Main Results:

• The proposed method provides a conservative adjustment for q-values.
• The adjustment method effectively accounts for FDR underestimation, even with permutation-based p-values.
• Demonstrated usefulness across diverse genomic datasets (microarray, sequencing) and simulations.

Conclusions:

• The mathematical conservativeness of the proposed approach is confirmed.
• Conservative adjustment of q-value is crucial for accurate FDR estimation.
• The method is particularly important when the proportion of differentially expressed genes is small or the differential expression signal is weak.