Abstract: The human microbiome is a complex assembly of bacteria that are sensitive to many perturbations. We have developed specific tools for studying the vaginal, intestinal and oral microbiomes under many using time course data following many different types of perturbations (pregnancy, hypo-salivation inducing medications and antibiotics are some examples).
A suite of statistical tools written in R and available as Bioconductor package (phyloseq, dada2) allows for easy denoising, normalization, visualization and statistical testing of the longitudinal multi-table data composed of 16S rRNA reads combined with clinical data, transcriptomic and metabolomic profiles. Challenges we have had to address include information leaks, the heterogeneity of the data, multiplicity of choices during the analyses and validation of results. A first step forward has been made through development of carefully designed denoising and normalization procedures, as the data themselves are not compositional, although the unknown parameters do belong to the simplex. Since the different taxa cannot be considered independent we have developed a statistical framework akin to Latent Dirichlet Allocation as it is used for textual analyses which are well adapted to the study of the stability of the human microbiome under antibiotic perturbations