Phylo-Spec: a phylogeny-fusion deep learning model advances microbiome status identification

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• The human microbiome plays a critical role in health and disease, offering potential for health state classification.
• Traditional machine learning (ML) and deep learning (DL) models for microbiome analysis often neglect microbial evolutionary relationships, leading to performance issues due to data sparsity and profiling inaccuracies.

Purpose of the Study:

• To introduce Phylo-Spec, a novel phylogeny-driven deep learning algorithm designed to improve microbiome-based health state classification.
• To address challenges in microbiome data analysis, including data sparsity, misclassified features, and unclassified species, by integrating multi-aspect microbial information.

Main Methods:

• Phylo-Spec fuses convolutional features of microbes within a phylogenetic hierarchy using a bottom-up iterative approach.
• The algorithm dynamically assigns unclassified species to virtual nodes on the phylogenetic tree and captures feature importance via an information gain mechanism.
• It integrates microbial abundance, taxonomy, and phylogeny within a unified deep learning framework.

Main Results:

• Phylo-Spec demonstrated superior efficacy in microbiome status classification on simulated and real-world metagenomic and amplicon datasets.
• The model significantly alleviated challenges associated with sparse data and inaccurate profiling compared to existing ML and DL methods.
• It successfully identified key microbial contributors linked to disease, enhancing classification interpretability.

Conclusions:

• Phylo-Spec establishes a powerful and interpretable framework for microbiome-based health state identification and microbe-disease association.
• The integration of phylogenetic information within a deep learning model offers a significant advancement over traditional microbiome analysis techniques.
• This approach holds promise for advancing microbiome-based diagnostics and precision medicine by leveraging evolutionary context.