Phylogenetic association analysis with conditional rank correlation

Area of Science:

• Microbiome Research
• Bioinformatics
• Statistical Genetics

Background:

• Phylogenetic association analysis is vital for microbiome studies.
• Existing methods struggle with high-dimensional data, linear assumptions, and confounding effects.
• There's a need for methods detecting complex, non-linear microbial associations.

Purpose of the Study:

• To introduce a novel phylogenetic association analysis framework.
• To address limitations of existing methods in handling complex associations and confounders.
• To develop robust tests for microbiome-outcome correlations.

Main Methods:

• Employed conditional rank correlation as the primary measure of association.
• Developed fully nonparametric tests to account for confounders, ensuring robustness.
• Utilized weighted sum and maximum approaches for aggregating subtree correlations; employed nearest-neighbor bootstrapping for significance calibration.

Main Results:

• The proposed framework successfully characterizes complex, non-linear associations in microbiome data.
• Nonparametric tests demonstrated robustness against outliers and effective handling of confounding variables.
• The bootstrapping method provided straightforward significance level determination and adaptability to new datasets.

Conclusions:

• The novel framework offers a powerful tool for microbiome association studies.
• It overcomes limitations of traditional methods, enabling deeper insights into microbial community functions.
• The approach is practical and validated on both simulated and real-world microbiome datasets.