Reconstructible phylogenetic networks: do not distinguish the indistinguishable

Area of Science:

• Evolutionary biology
• Phylogenetics
• Computational biology

Background:

• Phylogenetic networks model complex evolutionary histories including reticulate events like hybridization and gene transfer.
• Interpreting these networks involves analyzing the set of phylogenetic trees they display, representing potential evolutionary pathways.
• A key challenge is that different networks can display identical sets of trees, rendering them indistinguishable to many inference methods.

Purpose of the Study:

• To address the identifiability problem in phylogenetic network reconstruction.
• To propose a method for uniquely reconstructing phylogenetic networks from evolutionary data.
• To introduce a novel definition of uniquely reconstructible networks and a canonical representative for indistinguishable networks.

Main Methods:

• Investigated the "indistinguishability" problem where different phylogenetic networks yield the same set of displayed trees.
• Proposed that network inference should focus on uniquely identifiable structures.
• Introduced a definition for uniquely reconstructible networks and defined a canonical network for sets of indistinguishable networks.

Main Results:

• Identified that many phylogenetic networks are indistinguishable based on the trees they display.
• Demonstrated that accounting for branch lengths only partially resolves the identifiability issue.
• Proposed that only the canonical form of a phylogenetic network can be uniquely reconstructed from data exhibiting reticulate evolution.

Conclusions:

• The identifiability problem poses a fundamental limitation for reconstructing reticulate evolutionary histories.
• Unique reconstruction is only possible for the canonical form of a phylogenetic network.
• This necessitates a shift in perspective for interpreting phylogenetic networks and developing inference methods.