DisCo P-ad: Distance-Correlation-Based -Value Adjustment Enhances Multiple Testing Corrections for Metabolomics.

Area of Science:

• Biostatistics
• Bioinformatics
• Genomics
• Metabolomics

Background:

• High-throughput omics studies generate vast datasets with numerous noisy variables.
• Multiple testing correction is crucial to minimize false or missed discoveries in statistical analyses.
• Existing methods for multiple testing correction in omics, particularly metabolomics, may be too conservative, lenient, or limited in their assumptions.

Purpose of the Study:

• To propose a novel modification to p-value adjustment procedures for high-dimensional omics data.
• To enhance existing eigen-analysis-based multiple testing correction methods.
• To improve statistical power and reduce false positives in metabolome-wide association studies (MWAS).

Main Methods:

• Estimating the effective number of independent tests using eigen-analysis of the correlation matrix.
• Proposing a modification based on distance correlation, a more general measure of association.
• Evaluating the performance of the proposed method on metabolomics data with varying study designs.

Main Results:

• The distance correlation-based p-value adjustment (DisCo P-ad) demonstrated superior performance compared to existing methods.
• The study assessed common genome-wide association studies (GWAS) p-value adjustment procedures and one tailored for MWAS.
• Distance correlation showed improved results across various sample size-to-feature ratios, response types, and metabolite groupings.

Conclusions:

• DisCo P-ad offers a novel enhancement to eigen-analysis-based multiple testing correction.
• The method can increase statistical power and reduce false positives in omics studies.
• DisCo P-ad is applicable to metabolomics and other high-dimensional omics data analyses.