Compositionally Aware Phylogenetic Beta-Diversity Measures Better Resolve Microbiomes Associated with Phenotype.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbiome data present unique analytical challenges including sparsity and compositionality.
• Existing methods like RPCA, CTF, and UniFrac address some but not all of these challenges.
• Integrating prior information, such as phylogenetic structure and study design, is crucial for robust microbiome data analysis.

Purpose of the Study:

• To introduce novel methods, phylo-RPCA and phylo-CTF, that incorporate phylogenetic information into microbiome data analysis.
• To enhance the discriminatory power and classification accuracy of microbiome data analysis.
• To improve the effect size and robustness of pattern discovery in microbiome datasets.

Main Methods:

• Developed phylo-RPCA by integrating phylogenetic information into Robust Principal Component Analysis (RPCA).
• Developed phylo-CTF by integrating phylogenetic information into Compositional Tensor Factorization (CTF).
• Evaluated methods using simulated and real microbiome data, assessing discriminatory power, effect size, and classification accuracy.

Main Results:

• phylo-RPCA and phylo-CTF demonstrate substantial improvements over state-of-the-art methods.
• The addition of phylogenetic information quantitatively enhances effect size and classification accuracy.
• New methods effectively handle sparsity, compositionality, and phylogenetic structure in microbiome data.

Conclusions:

• phylo-RPCA and phylo-CTF offer significant advancements in microbiome data analysis.
• Incorporating phylogenetic information leads to more robust and accurate identification of microbial patterns.
• These methods improve the ability to distinguish between sample groups based on microbiome composition and evolutionary history.