[Paper Review] On Tree Based Phylogenetic Networks

Louxin Zhang|arXiv (Cornell University)|Sep 5, 2015

Genomics and Phylogenetic Studies9 references4 citations

TL;DR

This paper presents a simple necessary and sufficient condition for determining whether a phylogenetic network is tree-based, and constructs a universal tree-based network that displays every possible phylogenetic tree on a given set of species. The key contribution is proving that tree-based networks form a complete class, meaning any collection of trees on the same taxa can be simultaneously displayed in a single universal tree-based network.

ABSTRACT

A large class of phylogenetic networks can be obtained from trees by the addition of horizontal edges between the tree edges. These networks are called tree based networks. Reticulation-visible networks and child-sibling networks are all tree based. In this work, we present a simply necessary and sufficient condition for tree-based networks and prove that there is a universal tree based network for each set of species such that every phylogenetic tree on the same species is a base of this network. The existence of universal tree based network implies that for any given set of phylogenetic trees (resp. clusters) on the same species there exists a tree base network that display all of them.

Motivation & Objective

• To resolve two open problems posed by Francis and Steel (2015) regarding tree-based phylogenetic networks.
• To establish a simple, necessary and sufficient condition for a phylogenetic network to be tree-based.
• To construct a universal tree-based network $U_X$ that contains every phylogenetic tree on a set $X$ as a base.
• To demonstrate that the class of tree-based networks is complete, unlike reticulation-visible or galled networks.
• To support the use of tree-based networks as a natural and comprehensive model for horizontal gene transfer and reticulate evolution.

Proposed method

• Propose a new characterization of tree-based networks using vertex-disjoint path systems in a constructed universal network.
• Define a universal network $U$ with upper and lower components to embed all possible tree topologies.
• Use a one-to-one mapping $\pi$ to align leaf orders in a given tree with specific nodes in the universal network's lower component.
• Prove the existence of $m$ vertex-disjoint paths in the lower component $U_{\text{lower}}$ for any permutation $\pi$, ensuring no overlap in node usage.
• Construct the universal network $U_X$ recursively by embedding tree bases through path systems that preserve topology and avoid dummy vertices.
• Apply a linear-time algorithm to test tree-based status, enabling efficient analysis of network structures.

Experimental results

Research questions

• RQ1 What is a simple necessary and sufficient condition for a phylogenetic network to be tree-based?
• RQ2 Can a single phylogenetic network be constructed that displays every possible phylogenetic tree on a given set of species?
• RQ3 Is the class of tree-based networks complete, meaning it can represent any collection of trees on the same taxa?
• RQ4 How does the completeness of tree-based networks compare to that of reticulation-visible or galled networks?
• RQ5 What structural properties of the universal network ensure that all tree topologies can be embedded without overlap or dummy nodes?

Key findings

• A simple necessary and sufficient condition is established for a phylogenetic network to be tree-based, enabling efficient testing via a linear-time algorithm.
• A universal tree-based network $U_X$ exists for any set of species $X$, such that every phylogenetic tree on $X$ is a base of $U_X$.
• The universal network $U_X$ is constructed using vertex-disjoint path systems in its lower component, ensuring all tree topologies can be embedded without node conflicts.
• For any permutation $\pi$ of the leaf labels, $m$ vertex-disjoint paths exist in $U_{\text{lower}}$ connecting corresponding tree nodes, proving full embeddability.
• The class of tree-based networks is complete, in contrast to reticulation-visible networks, which are incomplete due to a bounded number of reticulation nodes.
• The construction confirms that tree-based networks are a natural and comprehensive model for horizontal gene transfer and other reticulate evolutionary processes.