MMETHANE: interpretable AI for predicting host status from microbial composition and metabolomics data.

Area of Science:

• Microbiome research
• Metabolomics
• Host-microbiome interactions
• Artificial Intelligence

Background:

• Metabolites are crucial for host health and microbiome interactions.
• Limited computational tools exist to link microbiome and metabolomic data to host status.
• Understanding these links is vital for disease diagnosis and treatment.

Purpose of the Study:

• To develop a computational tool for predicting host status using paired microbial and metabolomic data.
• To create an interpretable deep learning model that incorporates biological knowledge.
• To address limitations in current computational approaches for host-microbiome analysis.

Main Methods:

• Developed MMETHANE, a deep learning model integrating phylogenetic and chemical relationships.
• Trained and validated MMETHANE on six diverse datasets with paired microbial and metabolomic measurements.
• Utilized case studies on inflammatory bowel disease datasets to demonstrate biological insights.

Main Results:

• MMETHANE consistently performed on par with or outperformed existing methods across multiple datasets.
• The model demonstrated superior performance on 80% of evaluated datasets compared to other techniques.
• MMETHANE successfully identified biologically meaningful connections between microbes, metabolites, and disease states in case studies.

Conclusions:

• MMETHANE is an open-source, interpretable AI software package for microbiome research.
• The tool enhances the investigation of microbe-metabolite-host interplay, aiding disease mechanism understanding.
• MMETHANE facilitates improved diagnosis and treatment strategies for microbiome-impacted human diseases.