MultiMed

MultiMed implements a permutation-based testing framework to identify multiple biological mediators while controlling the family-wise error rate in studies with high-dimensional biomarkers such as gene expression or metabolite concentrations.

Key Features:

• Permutation methodology: Employs a permutation approach tailored for testing multiple mediators that focuses on the maximum of the test statistics to handle complex null hypotheses in datasets with hundreds or thousands of biomarkers.
• Control of family-wise error rate: Controls the family-wise error rate across multiple mediator tests to maintain rigorous type I error control.
• Improved statistical power: Demonstrates substantially greater power than Bonferroni correction in simulations, with reported 2–5× increases in power to detect true mediators even when biomarkers are independent.
• Empirical application: Applied to a case–control study of dietary risk factors and colorectal adenoma, identifying docosahexaenoate as a potential mediator between fish consumption and reduced colorectal adenoma risk among 149 tested metabolites.

Scientific Applications:

• Mediator discovery in omics studies: Detects putative mediators in high-dimensional gene expression and metabolomics datasets.
• Epidemiological mediation analysis: Tests multiple candidate mediators in biomedical and epidemiological study designs, including case–control studies.
• Pathway and mechanism inference: Supports inference of risk factor–disease pathways and investigation of pathophysiological processes through identified mediators.

Methodology:

Generates permutations of the data, computes test statistics using the maximum statistic to form a permutation-derived null distribution, and compares observed test statistics to that distribution to assess significance while controlling the family-wise error rate.