Detecting Association in Microbiome Compositional Data: A Novel Approach

In microbiome analysis, it is of interest to assess differences in compositions based on different phenotypes, with the fundamental goal of personalized care, providing each patient with the best possible therapy, tailored to their genes and phenotype. It is now established knowledge that the microbiome has a compositional nature, typically with an high-dimensional sparse structure, therefore it is necessary to identify suitable inferential procedures to not incur in inconsistent results. In this contribution we propose an heuristic approach based on a nonparametric independence test, capable of detecting nonlinear effects, to assess the compositional difference between many populations. The inferential problem is translated into testing the independence between a compositional variable and a categorical variable.