Computing all hybridization networks for multiple binary phylogenetic input trees.

Benjamin Albrecht

doi:10.1186/s12859-015-0660-7

Abstract

BackgroundThe computation of phylogenetic trees on the same set of species that are based on different orthologous genes can lead to incongruent trees. One possible explanation for this behavior are interspecific hybridization events recombining genes of different species. An important approach to analyze such events is the computation of hybridization networks.ResultsThis work presents the first algorithm computing the hybridization number as well as a set of representative hybridization networks for multiple binary phylogenetic input trees on the same set of taxa. To improve its practical runtime, we show how this algorithm can be parallelized. Moreover, we demonstrate the efficiency of the software Hybroscale, containing an implementation of our algorithm, by comparing it to PIRNv2.0, which is so far the best available software computing the exact hybridization number for multiple binary phylogenetic trees on the same set of taxa. The algorithm is part of the software Hybroscale, which was developed specifically for the investigation of hybridization networks including their computation and visualization. Hybroscale is freely available1 and runs on all three major operating systems.ConclusionOur simulation study indicates that our approach is on average 100 times faster than PIRNv2.0. Moreover, we show how Hybroscale improves the interpretation of the reported hybridization networks by adding certain features to its graphical representation.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-015-0660-7) contains supplementary material, which is available to authorized users.

Highlights

The computation of phylogenetic trees on the same set of species that are based on different orthologous genes can lead to incongruent trees
Results and discussion we first report a simulation study indicating that our approach is much faster than other existing methods and illustrate how Hybroscale can be used for studying hybridization networks by applying the software to a well known grass (Poaceae) dataset
Simulation study To show the efficiency of our implementation, we have integrated our algorithm into the Java software Hybroscale and conducted a simulation study comparing its runtime to PIRNv2.0 [7, 8], which is so far the best available software for computing exact hybridization numbers for multiple rooted binary phylogenetic X -trees

Summary

Introduction

The computation of phylogenetic trees on the same set of species that are based on different orthologous genes can lead to incongruent trees. Regarding a well-established homoploid hybrid species resulting from such a evolutionary process as described above, we can reconstruct its evolutionary history by taking two different scenarios each corresponding to one of its parental species into account This is typically done by, first, computing two rooted phylogenetic trees each based on those genes corresponding to one of both parental gene sets and, second, by reconciling these two topologically different trees into one rooted phylogenetic network, whose reticulate nodes (nodes of in-degree ≥ 2) represent certain putative hybridization events. Because of those major hurdles a hybrid species has to face, hybridization events rarely happen and, Albrecht BMC Bioinformatics (2015)16:236 from a biological point of view, only those networks containing a minimum number of reticulate nodes are of high interest

Methods

Results

Conclusion