Clustering Systems of Phylogenetic Networks

Abstract

Rooted acyclic graphs appear naturally when the phylogenetic relationship of a set $X$ of taxa involves not only speciations but also recombination, horizontal transfer, or hybridization, that cannot be captured by trees. A variety of classes of such networks have been discussed in the literature, including phylogenetic, level-1, tree-child, tree-based, galled tree, regular, or normal networks as models of different types of evolutionary processes. Clusters arise in models of phylogeny as the sets $\mathtt{C}(v)$ of descendant taxa of a vertex $v$. The clustering system $\mathscr{C}_N$ comprising the clusters of a network $N$ conveys key information on $N$ itself. In the special case of rooted phylogenetic trees, $T$ is uniquely determined by its clustering system $\mathscr{C}_T$. Although this is no longer true for networks in general, it is of interest to relate properties of $N$ and $\mathscr{C}_N$. Here, we systematically investigate the relationships of several well-studied classes of networks and their clustering systems. The main results are correspondences of classes of networks and clustering system of the following form: If $N$ is a network of type $\mathbb{X}$, then $\mathcal{C}_N$ satisfies $\mathbb{Y}$, and conversely if $\mathscr{C}$ is a clustering system satisfying $\mathbb{Y}$ then there is network $N$ of type $\mathbb{X}$ such that $\mathscr{C}\subseteq\mathscr{C}_N$.This, in turn, allows us to investigate the mutual dependencies between the distinct types of networks in much detail.