The inheritance procedure: multiple testing of tree-structured hypotheses.

Area of Science:

• Bioinformatics
• Statistical genetics
• Computational biology

Background:

• Hypothesis testing in bioinformatics often involves hierarchical data structures, such as probe, gene, and chromosome levels.
• Existing multiple testing procedures may not fully exploit these inherent graph structures, potentially limiting statistical power and interpretability.

Purpose of the Study:

• To introduce the inheritance procedure, a novel method for familywise error rate control in tree-structured hypothesis testing.
• To enhance statistical power and interpretability in bioinformatics analyses by utilizing hierarchical data relationships.

Main Methods:

• The inheritance procedure operates by initiating tests at the root of the hypothesis tree.
• Significant findings at higher levels trigger follow-up testing in relevant branches and neighboring leaf nodes.
• This approach systematically controls the familywise error rate across the entire hypothesis structure.

Main Results:

• The inheritance procedure demonstrates uniform improvement over the recently proposed method by Meinshausen.
• The method offers increased statistical power compared to non-hierarchical testing approaches.
• Results are more interpretable due to the structured approach to hypothesis testing.

Conclusions:

• The inheritance procedure provides an effective strategy for familywise error control in tree-structured bioinformatics data.
• Implementation within the globaltest package on Bioconductor facilitates its application in biological research.
• This method represents a significant advancement in analyzing complex, hierarchical biological data.