MicroBVS

MicroBVS implements Dirichlet-tree multinomial regression models with Bayesian variable selection in R to identify covariates associated with microbial taxa abundance while accounting for phylogenetic tree structure.

Key Features:

• Bayesian Variable Selection: Implements a Bayesian framework that evaluates posterior probabilities of covariate inclusion to select relevant predictors.
• Dirichlet-Tree Multinomial Models: Fits Dirichlet-tree multinomial regression models for compositional microbiome count data.
• Phylogenetic Structure Accommodation: Incorporates phylogenetic (tree) relationships among taxa within the Dirichlet-tree model structure.
• Flexible Covariate Parameterization: Supports various parameterizations of prior inclusion probabilities for covariates.
• Model Selection Uncertainty Consideration: Explicitly accounts for model selection uncertainty in inference on covariate effects.

Scientific Applications:

• Microbiome research: Identifies covariates associated with microbial taxa abundance in human microbiome studies.
• Compositional data analysis: Relates covariates to component abundances for compositional data with or without a known tree-like structure.

Methodology:

Constructs Dirichlet-tree multinomial models that incorporate phylogenetic relationships among taxa; uses a Bayesian approach to evaluate posterior probabilities of covariate inclusion, supports various prior inclusion parameterizations, and addresses model selection uncertainty.