A unified combination framework for dependent tests with applications to microbiome association studies

Area of Science:

• Statistics
• Microbiology
• Bioinformatics

Background:

• Meta-analysis traditionally combines independent studies.
• Combining dependent tests, common in microbiome research, poses statistical challenges.
• Existing methods like Cauchy combination have limitations with dependent data.

Purpose of the Study:

• Introduce a novel meta-analysis framework for combining dependent statistical tests.
• Generalize existing methods to rigorously handle dependence in microbiome association studies.
• Address limitations of current dependent test combination methods, including the Cauchy combination.

Main Methods:

• Developed a generalized meta-analysis framework for dependent tests.
• Built upon P-value aggregation and confidence distribution combination methods.
• Comprehensive simulation study comparing the proposed framework with existing dependent combination methods.

Main Results:

• The proposed framework provides rigorous statistical guarantees.
• Demonstrated that ignoring dependence can cause severe size distortion.
• The Cauchy combination method is a special case of the proposed framework.
• The framework effectively handles violations of distributional assumptions in Cauchy combination.
• Outperforms existing methods in terms of accurate size and enhanced power.

Conclusions:

• The novel framework accurately and efficiently combines dependent microbiome association tests.
• It offers flexibility and improved statistical power compared to existing methods.
• Enables more efficient and meaningful discoveries in microbiome research.