The tree-child network problem for line trees and the shortest common supersequences for permutations

TL;DR

This paper investigates the computational complexity of inferring minimal tree-child phylogenetic networks from multiple line trees, establishing NP-hardness and characterizing networks with minimal reticulate nodes.

Contribution

It proves NP-hardness of the tree-child network inference problem for line trees and characterizes minimal networks displaying all line trees and binary trees.

Findings

Tree-child network inference for line trees is NP-hard.

Minimal networks displaying all line trees have the same hybridization number as those displaying all binary trees.

Hybridization number for these networks is Θ(n^3) for n > 7 taxa.

Abstract

One strategy for reconstruction of phylogenetic networks is to solve the phylogenetic network problem, which involves inferring phylogenetic trees first and subsequently computing the smallest phylogenetic network that displays all the trees. This approach capitalizes on exceptional tools available for inferring phylogenetic trees from biomolecular sequences. Since the vast space of phylogenetic networks poses difficulties in obtaining comprehensive sampling, the researchers switch their attention to inferring tree-child networks from multiple phylogenetic trees, where in a tree-child network each non-leaf node must have at least one child that is an indegree-one node. Two results are obtained: (1) The tree-child network inference problem for multiple line trees remains NP-hard by a reduction from the shortest common supersequence problem for permutations and proving that the latter is NP-hard. (2) The tree-child networks with the least reticulate nodes that display all the line trees are the same as that display all the binary trees, whose hybridization number is $\Theta(n^3)$ for $n (> 7)$ taxa.