Closed testing with Globaltest, with application in metabolomics.

Area of Science:

• Metabolomics
• Statistical Genetics
• Bioinformatics

Background:

• The Globaltest is widely used for pathway testing in metabolomics to assess associations between feature sets and a response.
• Multiple testing correction is essential when evaluating numerous feature sets to avoid false discoveries.
• Existing methods may lack flexibility in post hoc analysis of feature sets.

Purpose of the Study:

• To develop a multiple testing procedure for the Globaltest that allows for post hoc selection of feature sets.
• To control the familywise error rate across all possible feature sets.
• To provide an efficient computational method for closed testing in metabolomics.

Main Methods:

• A closed testing-based multiple testing method tailored for the Globaltest.
• Derivation of a computational shortcut to overcome the exponential complexity of traditional closed testing.
• Implementation of the shortcut procedure in an R package (ctgt).

Main Results:

• The proposed method controls the familywise error rate simultaneously for all feature sets.
• The shortcut enables exact closed testing on metabolomics data, significantly reducing computation time.
• Demonstrated application of the method on real metabolomics datasets.

Conclusions:

• The developed method offers a statistically rigorous and computationally efficient approach for multiple testing in Globaltest applications.
• Researchers can now select feature sets of interest post-data analysis without compromising error control.
• The ctgt R package facilitates the practical application of this advanced statistical technique in metabolomics research.