Figure 4.

Bayesian networks of correlations of the bacterial OTUs. Positive (red arcs) and negative (blue arcs) correlations (Spearman correlation >0.3 and < −0.3) between bacterial OTUs visualized as a Bayesian network using R package “bnlearn.” The connecting arcs of the network represented mutual associations (not causality). Separate networks were constructed for all samples combined together (A) and also separated according to the experimental group (B). Separate networks for each experimental group in the (B) contain also correlations with mice body weight and experimental time. The bold arcs in the (B) connect the bacterial OTUs with have the most important associations to the body weight and experimental time (so called dissection of Markov blankets). VehD.group was excluded from the analysis due to the low number of time-points.