Improving intraspecific allele networks inferred by maximum parsimony

Abstract

Allele (or haplotype) networks are often used in phylogeographic studies to display genetic variation within a species or a group of closely related species. A global maximum parsimony approach to infer allele networks, arguably the method of choice to display genetic variation at the intraspecific level, consists in inferring all most parsimonious trees from a DNA sequence alignment and combining the corresponding phylograms into a single graph. However, it has been suggested that, while classic phylogenetic programs generate a single phylogram per most parsimonious tree, deriving all possible phylograms from them would allow identifying additional most parsimonious paths among alleles, thereby improving this network inference method. We test this prediction by analysing both simulated and empirical DNA sequence alignments. For this purpose, a computer program, CPN, was developed to implement the entire procedure, starting with a set of most parsimonious trees and combining all derived phylograms into a network. We show that including all possible most parsimonious phylograms indeed often results in finding additional most parsimonious paths in the network graph, thereby improving the search for a global maximum parsimony solution. We highly recommend the use of this approach in future phylogeographic studies, to ensure that all most parsimonious paths are included in the allele network, instead of an arbitrarily selected subset of those.