[Paper Review] Tripartitions do not always discriminate phylogenetic networks

Gabriel Cardona, Francesc Rosselló+1 more|ArXiv.org|Jul 16, 2007

Genomics and Phylogenetic Studies7 references3 citations

TL;DR

This paper demonstrates that the tripartition metric—proposed by Moret, Nakhleh, and Warnow for comparing phylogenetic networks—fails to distinguish non-isomorphic networks in all subclasses where it was claimed to be effective. The authors provide counterexamples showing that distinct networks can share identical tripartition sets, invalidating the metric's separation axiom, and identify a restricted subclass (tree-child, weakly time-consistent networks) where tripartitions and even bipartitions do enable a valid metric.

ABSTRACT

Phylogenetic networks are a generalization of phylogenetic trees that allow for the representation of non-treelike evolutionary events, like recombination, hybridization, or lateral gene transfer. In a recent series of papers devoted to the study of reconstructibility of phylogenetic networks, Moret, Nakhleh, Warnow and collaborators introduced the so-called {tripartition metric for phylogenetic networks. In this paper we show that, in fact, this tripartition metric does not satisfy the separation axiom of distances (zero distance means isomorphism, or, in a more relaxed version, zero distance means indistinguishability in some specific sense) in any of the subclasses of phylogenetic networks where it is claimed to do so. We also present a subclass of phylogenetic networks whose members can be singled out by means of their sets of tripartitions (or even clusters), and hence where the latter can be used to define a meaningful metric.

Motivation & Objective

• To evaluate the discriminating power of the tripartition metric in phylogenetic network comparison.
• To challenge the claim that tripartitions uniquely identify networks in proposed subclasses, such as tree-sibling and strongly time-consistent networks.
• To identify a subclass of phylogenetic networks where tripartitions (or bipartitions) do satisfy the separation axiom and thus can define a meaningful metric.
• To clarify the topological conditions under which tripartitions truly reflect network structure.
• To guide future development of robust metrics for phylogenetic network comparison beyond the flawed tripartition approach.

Proposed method

• Construct explicit counterexamples of non-isomorphic phylogenetic networks that share identical sets of tripartitions.
• Analyze the structure of tripartitions by partitioning leaves into strict descendants, non-strict descendants, and non-descendants of each node.
• Use graph-theoretic reasoning to show that isomorphism cannot be guaranteed even when tripartition sets are equal.
• Prove that in tree-child, weakly time-consistent networks, tripartitions and bipartitions (via Bourque-Robinson-Foulds) do uniquely identify networks.
• Establish a bijective mapping between clusters and tripartitions in the valid subclass, ensuring that cluster equality implies isomorphism.
• Apply the separation axiom to define a proper distance metric on the corrected subclass using symmetric difference of tripartition sets.

Experimental results

Research questions

• RQ1 Do tripartitions uniquely determine the isomorphism class of phylogenetic networks in the subclasses proposed by Moret et al.?
• RQ2 Can non-isomorphic phylogenetic networks have identical sets of tripartitions, thereby invalidating the tripartition metric?
• RQ3 What structural constraints on phylogenetic networks ensure that tripartitions or bipartitions can serve as a basis for a valid distance metric?
• RQ4 Is there a well-defined subclass of phylogenetic networks where equality of tripartition multisets implies isomorphism?
• RQ5 What topological features of networks lead to indistinguishability under tripartition-based metrics?

Key findings

• The tripartition metric fails to satisfy the separation axiom in all subclasses where it was claimed to be effective, including tree-sibling and strongly time-consistent networks.
• Counterexamples exist where non-isomorphic phylogenetic networks have identical sets of tripartitions, proving the metric cannot distinguish them.
• In tree-child, weakly time-consistent phylogenetic networks, tripartitions uniquely determine the network structure and thus can define a valid metric.
• Even bipartitions in the sense of Bourque-Robinson-Foulds can serve as a basis for a meaningful distance in this subclass.
• The authors identify a well-defined subclass where cluster-based representations (via tripartitions) are sufficient to ensure isomorphism.
• The paper establishes that the tripartition metric cannot be used reliably for network comparison in the classes studied by Moret et al., necessitating a reevaluation of network distance metrics.