Robust differential abundance test in compositional data

Summary Differential abundance tests for compositional data are essential and fundamental in various biomedical applications, such as single-cell, bulk RNA-seq microbiome analysis. However, because of the constraint prevalence zero counts data, differential analysis on remains a complicated unsolved statistical problem. This article proposes new test, robust to address these challenges. Compared with existing methods, test is simple computationally efficient, prevalent datasets, can take data’s nature into account, has theoretical guarantee controlling false discoveries general setting. Furthermore, presence observed covariates, work covariate-balancing techniques remove potential confounding effects draw reliable conclusions. The proposed applied several numerical examples, its merits demonstrated using both simulated real datasets.