Fig. 1

Chart illustrating the three modules of the phyloMDA package proposed here for phylogeny-aware microbiome data analysis. phyloMDA requires three input files: a count matrix, a metadata matrix, and a phylogenetic tree. Module I contains R functions for fitting (zero-inflated) Dirichlet-tree multinomial models for multivariate abundance data. Some of these functions are invoked in Module II to produce tree-guided empirical Bayes (eBay) estimates of microbial compositions. These relative abundances are then used as input into high-level analyses. In particular, Module III contains R functions for tree-based multiscale regressions with relative abundances as predictors