DeepPhylo: Phylogeny-Aware Microbial Embeddings Enhanced Predictive Accuracy in Human Microbiome Data Analysis

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbiome data analysis is complex due to high dimensionality, sparsity, and compositionality.
• Integrating microbial abundance and phylogenetic information improves pattern discovery and predictive performance.
• Existing methods often neglect evolutionary distances within phylogenetic trees.

Purpose of the Study:

• To introduce DeepPhylo, a novel method for microbiome data analysis.
• To effectively integrate microbial abundance and phylogenetic information using phylogeny-aware amplicon embeddings.
• To enhance both unsupervised discriminatory power and supervised predictive accuracy in microbiome studies.

Main Methods:

• Development of DeepPhylo, a novel computational method.
• Utilizing phylogeny-aware amplicon embeddings to capture evolutionary distances.
• Integration of abundance and phylogenetic information for comprehensive analysis.

Main Results:

• DeepPhylo demonstrated superior performance compared to existing methods.
• The method improved unsupervised discriminatory power and supervised predictive accuracy.
• Biologically relevant insights were enhanced across diverse microbiome applications.

Conclusions:

• DeepPhylo offers a superior approach to microbiome data analysis by incorporating evolutionary distances.
• The method shows significant improvements in biological insight generation and predictive tasks.
• DeepPhylo is effective across multiple real-world microbiome use cases, including disease diagnosis and host characteristic prediction.