Phylogeny-based Principal Component Analysis.

<p>Bacterial communities of the three different treatment groups were clustered using PCA and the unweighted UniFrac distance matrix as an input (weighted PCA is shown in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0059260#pone.0059260.s006" target="_blank">Figure S6</a>). With PCA, a multivariate statistical analyses, axes that reflect the largest part of sample variation are identified (Percentage values at the axes reflect the level of variation explained by each principal coordinate; the first axis indicates the largest fraction of difference). Separation of the different sample collectives in 3 dimensions is visualized. A separation of the samples from the intervention group (I), that is most predominant 4 weeks after smoking cessation, was revealed. In contrast, the samples from the non-smoking (N) and smoking (S) control groups clustered together closely, thus reflecting their overall similar microbial composition.