SMARTscan

SMARTscan identifies taxonomic groups in microbiome sequence data that are associated with clinical traits and outcomes using a tree-based variable grouping and hierarchical model selection approach.

Key Features:

• Taxonomy-Based Grouping: Organizes potential predictors into hierarchical groups based on a predefined taxonomy tree.
• Hierarchical Model Selection: Performs a hierarchical search and determination of variable groups for association testing using taxonomic information.
• Increased Analytical Power: Improves analytical power relative to single-variable analysis when taxonomic group effects are present.
• Comparative Performance: Outperforms stepwise regression, LASSO (Least Absolute Shrinkage and Selection Operator), and CART (Classification and Regression Tree) in simulations.

Scientific Applications:

• Disease and Trait Association Discovery: Identifies taxonomic groups associated with diseases or phenotypic traits in microbiome datasets.
• Empirical Case Study: Applied to a vervet monkey diet experiment to detect phenotype-associated taxonomic features missed by other methods.

Methodology:

Organizes microbiome sequence data into optimized taxonomic groups using a predefined taxonomy tree, then hierarchically searches these groups to identify associations with clinical traits or diseases, with performance evaluated by simulations against single-variable analysis, stepwise regression, LASSO, and CART.