Compositional Mediation Analysis for Microbiome Studies

Motivated by recent advances in causal inference on mediation analysis and problems in the analysis of metagenomic data, we consider the effect of a treatment on an outcome transmitted through microbes, or compositional mediators. Compositional and high dimensional natures of such mediators make the standard mediation analysis not directly applicable. In this paper, we propose a method for estimating the expected causal direct and indirect (or mediation) effects for a sparse high-dimensional compositional mediation model utilizing the algebra for compositions in the simplex space and the nature of compositional data. The test of the total and component-wise mediation effects are also proposed. We conduct extensive simulation studies to assess the performance of the proposed method and apply the method to a real metagenomic dataset to investigate the effect of fat intake on body mass index (BMI) mediated through the gut microbiome composition. . CC-BY 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/149419 doi: bioRxiv preprint first posted online Jun. 13, 2017;