A Simultaneous Feature Selection and Compositional Association Test for Detecting Sparse Associations in...

Area of Science:

• Microbiome research
• Metagenomics
• Statistical bioinformatics

Background:

• Metagenomic studies often seek microbial associations with phenotypes.
• Multivariate analysis without prior knowledge can reduce statistical power.
• High dimensionality and compositional constraints challenge microbiome association discovery.

Purpose of the Study:

• To develop a novel method for robust microbial community structure analysis.
• To address limitations in statistical power and generalizability in metagenomic association studies.
• To introduce the Selection-Energy-Permutation (SelEnergyPerm) method.

Main Methods:

• Developed the nonparametric Selection-Energy-Permutation (SelEnergyPerm) method.
• Integrated feature selection with logratio signatures to account for compositional data.
• Employed simulations to evaluate performance across various association signal densities.

Main Results:

• SelEnergyPerm identifies parsimonious logratio sets capturing strong associations.
• The method demonstrates consistent detection/rejection of associations in synthetic data.
• Case studies using 16S and whole-genome sequencing data showcase method benefits.

Conclusions:

• SelEnergyPerm offers a powerful, statistically robust approach for microbiome association studies.
• The method effectively handles compositional data and sparse association signals.
• An R implementation is available for broader application in microbial ecology research.