Coracle-a machine learning framework to identify bacteria associated with continuous variables.

Area of Science:

• Microbiology
• Bioinformatics
• Artificial Intelligence

Background:

• Microbial communities play crucial roles in host health and environmental adaptation.
• Identifying specific microbial taxa associated with host traits is essential for understanding these roles.
• Current methods may struggle with the complexity and scale of microbiome data.

Purpose of the Study:

• To introduce Coracle, a novel artificial intelligence (AI) framework for identifying associations between bacterial communities and continuous host variables.
• To provide a robust and scalable tool for microbiome data analysis.
• To demonstrate Coracle's utility in uncovering bacterial associations with coral thermal tolerance.

Main Methods:

• Coracle employs an ensemble approach combining feature selection methods and machine learning (ML) models.
• It aggregates results into a score based on ML model performance, feature importance, and selection robustness.
• Regression coefficients offer initial insights into the direction of associations.

Main Results:

• Coracle successfully identified bacterial taxa associated with coral thermal tolerance, including previously known ones.
• The framework demonstrates scalability with large datasets, handling hundreds to thousands of features.
• Running analyses at higher taxonomic levels first (e.g., order, family) is recommended for efficiency.

Conclusions:

• Coracle is an effective AI tool for discovering microbial associations with host phenotypes.
• The framework offers a scalable and robust solution for analyzing complex microbiome datasets.
• Coracle can be accessed via a web server and its code is open-source.