A0362
Title: DisCo P-ad: Distance Correlation-based P-value Adjustment boosts multiple-testing corrections in metabolomics analyses
Authors: Debmalya Nandy - Center for Innovative Design & Analysis, Colorado School of Public Health, University of Colorado Anschutz Medical Campus (United States) [presenting]
Debashis Ghosh - University of Colorado Anschutz Medical Campus (United States)
Katerina Kechris - Colorado School of Public Health (United States)
Abstract: High-throughput data, often encountered in -omics sciences (e.g., genomics, metabolomics), contain measurements on several hundred or thousands of variables. In tests of association of these predictors with a clinical outcome of interest, multiple-testing corrections mitigate the number of false and truly missed discoveries. Many corrections involve first estimating the effective number of tests (number of statistically independent predictors among all original ones) for a subsequent Bonferroni-type adjustment to obtain the point-wise significance level, corresponding to a preset overall type-I error rate. Such practice is commonplace in Genome-Wide Association Studies (GWAS) but is also relevant to Metabolome-Wide Association Studies (MWAS). For MWAS, we consider procedures for p-value adjustments in GWAS along with one specifically designed for MWAS. While most are based on eigen-analysis of the Pearson's correlation matrix of the predictors, we propose using the Distance Correlation instead in the eigen-analysis for P-value adjustment (DisCo P-ad). Our extensive simulation study, based on real metabolomics datasets, demonstrates superior performance of DisCo for varying sample sizes, nature of the response (continuous/binary), and groupings of the metabolites. In summary, our study introduces a new method for enhancing multiple testing corrections in MWAS.
