SelEnergyPerm

SelEnergyPerm detects sparse compositional associations between microbial taxa and phenotypes in high-dimensional metagenomic data using logratio-based, nonparametric group association testing with embedded feature selection.

Key Features:

• Simultaneous feature selection and association testing: Integrates feature selection with a group-level association test to identify sparse associations among taxa in high-dimensional data.
• Nonparametric group association test: Employs a nonparametric approach to assess group-level associations without relying on parametric distributional assumptions.
• Embedded feature selection of logratios: Identifies small, independent sets of logratios between taxonomic features to produce parsimonious signatures.
• Compositional data handling: Uses logratios to account directly for the compositional constraints of metagenomic count data.
• Robustness across scenarios: Demonstrates consistent detection or rejection of associations in simulations with sparse, dense, or no association signals.
• Addresses low sample size and power limitations: Enhances statistical power in settings with limited sample size by focusing on parsimonious logratio signatures.

Scientific Applications:

• Microbial-phenotype association discovery: Detects associations between microbial compositions and phenotypes in environments such as gut, skin, and oral microbiomes.
• 16S amplicon sequencing analysis: Applied to publicly available 16S amplicon datasets to identify compositional signatures linked to phenotypes.
• Whole-genome sequencing (WGS) analysis: Applied to whole-genome sequencing datasets for characterization of microbial community alterations across environments.
• Community structure characterization: Identifies parsimonious logratio signatures that describe alterations in microbial community structure.

Methodology:

Embeds feature selection within a compositional association test framework using logratios and a nonparametric group association test to identify parsimonious logratio signatures for dimensionality reduction and association assessment.