Compositional analysis of microbiome data using the linear decomposition model (LDM).

Abstract

There are compelling reasons to test compositional hypotheses about microbiome data. We present here linear decomposition model-centered log ratio (LDM-clr), an extension of our LDM approach to allow fitting linear models to centered-log-ratio-transformed taxa count data. As LDM-clr is implemented within the existing LDM program, this extension enjoys all the features supported by LDM, including a compositional analysis of differential abundance at both the taxon and community levels, while allowing for a wide range of covariates and study designs for either association or mediation analysis. LDM-clr has been added to the R package LDM, which is available on GitHub at https://github.com/yijuanhu/LDM.