TSLC: differential abundance analysis method for identifying biologically relevant known microbes

Identifying biologically relevant microbes is a crucial task in human microbiome studies. Numerous computational and statistical methods have been proposed for this task. However, most existing methods are based solely on taxonomic information, even when whole-metagenome sequencing data are available, and thus cannot account for the fact that different microbes can perform the same or similar functions (known as functional redundancy) and for the contribution of unknown microbes to functional abundance. Hence, the identified differentially abundant microbes might not be biologically relevant to manifested phenotypes.

To address the issues with existing methods, we propose a two-stage procedure: (i) determining differentially abundant functions between groups; and then (ii) determining microbes significantly associated with each identified differentially abundant function. This procedure involves ${\ell }_1$ -penalized log-contrast models and debiasing procedures to account for the compositional nature of microbiome data and the bias induced by the ${\ell }_1$ penalty. We performed extensive simulation studies to illustrate the consequences of ignoring functional redundancy and the contribution of unknown microbes, and to assess the performance of the proposed method. We applied the proposed method to two whole-metagenome sequencing datasets and showcased substantial differences between the typical taxonomic-based analysis and the proposed method.

An R package for the proposed method is available at https://github.com/rgbeblavy/TSLC.