SplitsTree

SplitsTree4 – a Java Framework
for Phylogenetic Trees and
Networks
Daniel Huson
www-ab.informatik.uni-tuebingen.detuebingen.de

Copyright (c) 2008 Daniel Huson.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.2
or any later version published by the Free Software Foundation;
with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
Texts. A copy of the license can be found at http://www.gnu.org/copyleft/fdl.html

Trees vs networks
• Evolutionary relationships are usually
represented by phylogenetic trees
• But: real data contain different and/or
conflicting signals, and thus do not always
clearly support a unique tree
• Enter phylogenetic networks, as either:
• simply a visualization of conflicting data, or
• a more complex model of evolution containing
events such as recombination or hybridization

Trees vs Networks
Domain
Bacteria
Eukaryotes
Archaea
Bacteria
Eukaryotes
Archaea
Kingdom
Proteobacteria
Cyanobacteria
Animals
Fungi
Plants
Archezoa
Euryarchacota
Crenarchaeota
Proteobacteria
Cyanobacteria
Animals
Fungi
Plants
Archezoa
Euryarchacota
Crenarchaeota
Doolittle, 1999
Doolittle, 1999
Tree of life
Web of life

Computed using
Neighbor-Joining
Trees vs Networks
Computed using
split decomposition
Neisseria phylogeny (Eddie Holmes, 1999)

Trees and splits
The split encoding (T) of a tree T:
G 6
G 1
G 4
G 8
e
G 5
G 2
G 7
G 3
G 1 ,G 3 ,G 4 ,G 6 ,G 7 vs G 2 ,G 5 ,G 8

Networks and splits
Cut-set of parallel edges defines split { {A,B}} vs rest

Splits and splits graphs
• Any given system of splits can be
represented by a splits graph G. G
Note that:
• G is a tree iff is compatible
(e.g. Neighbor-Joining)
• G is outer- labeled- planar iff is circular
(e.g. Neighbor-Net, Net, Bryant & Moulton 2002)
• G is usually planar or only mildy non-planar iff
is weakly compatible (e.g. Split Decomposition)
• G is always subgraph of n-dim. hypercube
(e.g. recoding of sequences, spectral analysis, median networks,
consensus networks, Z-super networks)
(Theory of splits worked out by Bandelt and Dress 1992)

SplitsTree 3. 2
Implements split
decomposition and
related methods
First version developed
with Rainer Wetzel in
1995
Current version 3.2 in
C++ using Tcl-TkTk
Runs under Linux, Unix,
Windows and MacOS

Design criteria for SplitsTree4
• Must run on any machine with minimal
installation requirements
• GUI for interactive use, command- line for
pipelines
• Open system, decentralized plug-in concept
• Based on splits, also including quartets etc
• Based on Nexus file format, with support
for most common formats

Data flow in SplitsTree
Taxa
Taxa are
represented e.g. by
aligned sequences
Unaligned
Assumptions
Characters
Bootstrap
Transform
characters into
distances e.g. using
Hamming Transform distances
splits in
Transform
to unrooted or
Every connector
distances into splits
rooted graph or tree
represents a e.g. data
using Neighbor-
Analysis
transformation
net
(plug-in)
Distances
Quartets
Trees
Splits
Graph

Writing a new transformation
A new tree-building method “BestTree” is provided to
SplitsTree as follows:
pub l i c c l a s s Be s t T r e e i mp l emen t s D i s t anc e s2T r e e
{
/ / r e t u r ns t r ue , i f Be s t T r e e i s app l i c ab l e
pub l i c boo l e an i sApp l i c ab l e ( Ta x a t , D i s t anc e s d )
{ … }
/ / app l i e s Be s t T r e e and r e t u r ns t he t r e e
pub l i c T r e e app l y ( Ta x a t , D i s t anc e s d ) { … }
/ / commun i c a t i ng op t i ons t o Sp l i t sT r e e :
i n t ge t Op t i onTh r e sho l d ( ) { … }
vo i d s e t Op t i onTh r e sho l d ( i n t t ) { … }
}

SplitsTree Windows

SplitsTree Windows
Taxa
Unaligned
Characters
Distances
Quartets
Trees
Splits
Main Window
Method Window

SplitsTree Editor

Individual Gene Trees
ITS00
46 taxa

Individual Gene Trees
ITS03
40 taxa

Individual Gene Trees
SSU00
29 taxa

Individual Gene Trees
SSU03
40 taxa

Individual Gene Trees
Gpd03
40 taxa

Gene Trees as Super Network

Gene Trees as Super Network
ITS00+
ITS03

Gene Trees as Super Network
ITS03+
SSU00

Gene Trees as Super Network
ITS00+
ITS00+
SSU03

Gene Trees as Super Network
ITS00+
ITS03+
SSU03+
Gpd03

Gene Trees as Super Network
ITS00+
ITS03+
SSU00+
SSU03+
Gpd03

Exponential Explosion
Methods like the consensus network,
Z-super network or bootstrap-network
only produce a polynomial number of
splits
The number of nodes and edges of the
corresponding splits graph can grow
exponentially…
How to deal with this

Incompatibility Graph IG()
Nodes: splits
Edges: pairs of incompatible splits
Note: A 3-cube in the
splits graph corresponds
to a 3-clique in IG( )

All splits of 50 Gene Trees on Archaea
D=2
D=3
D=4
D=5
D=10

Hybridization Networks
Currently developing algorithms for
computing hybridization networks and
ancestor recombination graphs

Summary
SplitsTree provides a frame-work for
phylogenetic analysis
Extensibility based on plug-in design
Built on splits, incorperates both tree and
network methods
Provides all popular distance-based tree
building algorithms
Provides network methods such as split
decomposition, Neighbor-net, consensus
networks and super networks.

Credits
Authors: David Bryant and D.H.
Additional programming:
Tobias Dezulian, Markus Franz,
Miguel Jette,Tobias Kloepper, and
Michael Schröder