dICC: distance-based intraclass correlation coefficient for metagenomic reproducibility studies.

Area of Science:

• Microbiology
• Bioinformatics
• Biostatistics

Background:

• Microbiome data analysis often relies on pairwise distances due to sparsity and high dimensionality.
• Reproducibility of microbiome sampling methods is crucial for statistical power and biological insights.
• Traditional intraclass correlation coefficient (ICC) is used for univariate measurements but not directly applicable to distance matrices.

Purpose of the Study:

• To extend the traditional intraclass correlation coefficient (ICC) to microbiome distance measures.
• To propose a novel distance-based ICC (dICC) for quantifying reproducibility in microbiome studies.
• To enable statistical inference for microbiome data reproducibility using the derived asymptotic distribution of dICC.

Main Methods:

• Extension of the traditional intraclass correlation coefficient (ICC) to distance metrics.
• Development of a distance-based ICC (dICC) tailored for microbiome data.
• Derivation of the asymptotic distribution for the sample-based dICC to support statistical inference.

Main Results:

• A new metric, distance-based ICC (dICC), is proposed for microbiome reproducibility.
• The asymptotic distribution of dICC is derived, allowing for statistical inference.
• The utility of dICC is demonstrated using a real-world metagenomic reproducibility study.

Conclusions:

• The proposed distance-based ICC (dICC) offers a robust method for assessing microbiome reproducibility.
• dICC facilitates more powerful statistical analyses of microbiome distance matrices.
• The R package GUniFrac implements dICC for practical application in microbiome research.