[Paper Review] The compressions of reticulation-visible networks are tree-child

Andreas D. M. Gunawan, Hongwei Yan+1 more|arXiv (Cornell University)|Jun 20, 2018

Genomics and Phylogenetic StudiesBiochemistry, Genetics and Molecular Biology1 references4 citations

TL;DR

This paper introduces a rigorous compression operation for rooted phylogenetic networks, proving that compressing a reticulation-visible network yields a tree-child network. The method enables a linear-time algorithm for the cluster containment problem in a new class of networks called quasi-reticulation-visible networks, establishing a novel connection between reticulation-visible and tree-child networks.

ABSTRACT

Rooted phylogenetic networks are rooted acyclic digraphs. They are used to model complex evolution where hybridization, recombination and other reticulation events play important roles. A rigorous definition of network compression is introduced on the basis of the recent studies of the relationships between cluster, tree and rooted phylogenetic network. The concept reveals another interesting connection between the two well-studied network classes|tree-child networks and reticulation-visible networks|and enables us to define a new class of networks for which the cluster containment problem has a linear-time algorithm.

Motivation & Objective

• To formally define a compression operation for rooted phylogenetic networks (RPNs) based on structural simplification of non-leaf, non-reticulate nodes.
• To investigate the closure properties of network classes under compression, particularly focusing on reticulation-visible and tree-child networks.
• To introduce a new class of networks—quasi-reticulation-visible networks—by extending the concept of reticulation visibility.
• To demonstrate that the cluster containment problem remains solvable in linear time for this new network class.
• To establish a deeper structural connection between reticulation-visible and tree-child networks through compression.

Proposed method

• Define network compression as replacing connected components of non-leaf, non-reticulate nodes (including tree nodes and redundant nodes) with a single representative node.
• Use node coloring (red, blue, purple) in the compressed network to track visibility and dominance relationships in the original network.
• Prove that if the original network is reticulation-visible, its compression is tree-child by verifying that every non-leaf node has a tree-child.
• Construct a spanning tree in the compressed network to verify tree-child properties and leaf visibility.
• Introduce the concept of quasi-reticulation-visible networks by relaxing visibility constraints while preserving linear-time solvability of the cluster containment problem.
• Design a linear-time algorithm for the cluster containment problem in quasi-reticulation-visible networks using visibility and dominance checks in the compressed network.

Experimental results

Research questions

• RQ1 Does the compression of a reticulation-visible network result in a tree-child network?
• RQ2 Can a new class of networks be defined such that the cluster containment problem remains solvable in linear time?
• RQ3 Are tree-sibling networks closed under compression, and what are the implications for network reconstruction?
• RQ4 How does the compression operation relate to the visibility of nodes and the display of clusters in the network?
• RQ5 Can the compression framework be used to unify or compare different classes of phylogenetic networks?

Key findings

• The compression of any reticulation-visible network is guaranteed to be a tree-child network, establishing a new structural link between these two network classes.
• The cluster containment problem can be solved in linear time for a newly defined class of networks called quasi-reticulation-visible networks.
• Tree-sibling networks are not closed under compression, indicating a fundamental limitation of this network class in terms of structural stability under simplification.
• The compression operation preserves visibility and dominance relationships, allowing the reconstruction of leaf display relationships in the original network from the compressed version.
• The visibility of a node in the original network corresponds to the presence of a path from its representative in the compressed network to a leaf, enabling efficient cluster containment checks.
• The method enables a linear-time algorithm for the cluster containment problem in quasi-reticulation-visible networks by verifying three structural conditions in the compressed network.