(http://d3js.org). There are already many excellent phylogeny viewers out
there, this project is an experiment to see how well-suited the unique dynamic
interactive features of d3 are to phylogeny visualization.
## Example use
To view the siphonophore phylogeny:
python -m SimpleHTTPServer 8888 &
Then point your browser to http://localhost:8888/force.html. Click and drag the
nodes, mouse-over to see the names.
The included sample tree, siphonophorae.json, has 191 nodes (internal and
external) and works well in chrome and safari. Here is how you would generate
the siphonophore tree from scratch, as well as some larger trees (they are
increasingly inclusive subgraphs that include siphonophores):
curl http://www.itis.gov/downloads/itisMySQLTables.tar.gz | tar zx
cd [path to phylogeny-d3...]
# In the command below, use the full paths to the files you downloaded above
# 191 node tree:
python itis_sql_to_json.py taxonomic_units strippedauthor 718928 > force/siphonophorae.json
# 2437 node tree:
python itis_sql_to_json.py taxonomic_units strippedauthor 48739 > force/hydrozoa.json
# 6240 node tree:
python itis_sql_to_json.py taxonomic_units strippedauthor 48738 > force/cnidaria.json
# 366809 node tree:
python itis_sql_to_json.py taxonomic_units strippedauthor 202423 > animalia.json
To examine these trees, replace 'siphonophorae.json' in force/force.html with
the name of the new file generated above. Also, don't forget to chmod 755 your
itis_sql_to_json.py - parses taxonomic data from itis.gov into a tabular tree
encoded in json. Each node has an associated date. For tips, the date is the
year in which the species was described. For internal nodes, it is the year
that the youngest descendent species was described. The json file is in a
nodes, links format. The itis taxonomy is not a phylogeny (it describes the
categorization of animals, not their evolutionary relationships). But it is the
simplest way to get a tree with hundreds of thousand of tips for working on
large visualization projects.
force - a d3 tree viewer. It can read the json files, including those produced
### The anatomy of a phylogeny
A phylogeny is a depictions of the evolutionary relationships, such as a family
tree for a group of species or DNA sequences.
A phylogeny can be unrooted (ie, an undirected acyclic graph), in which case it
isn't indicated which node is the oldest and to assertion about the direction
of time is made on any edge, or rooted (ie, a DAG), in which case one node is
declared to be the root ancestor and time proceeds along edges from the root to
Edge length can have no meaning (a cladogram), indicate the amount of observed
or expected evolutionary time (a phylogram), or be proportional to time (a
Sometimes the topology of the tree is the main thing of interest. Sometimes the
topology is the backbone for understanding other data that are mapped onto the
### Data structures
The most common phylogeny datastructure is the Newick format:
It is also common to store trees as simple graph tables. There can be separate
tables for nodes and edges, or one table for nodes that includes a field that
specifies the parent node.
NeXML (http://www.nexml.org) is a promising standard for representing
phylogenetic data, including trees.
### Existing phylogeny viewers
There are already many excellent tools for viewing phylogenies. Most of these
tools are intended to render static snapshots, and interactivity is usually
focused on optimizing the layout rather than active data exploration.
#### Existing stand-alone viewers
FigTree (http://tree.bio.ed.ac.uk/software/figtree/) - the primary workhorse
used by the scientific community used for drawing trees for publication. It
enables basic interaction, such as toggling
Dendroscope (http://ab.inf.uni-tuebingen.de/software/dendroscope/) is
particularly well suited for very large trees with many tips.
Mesquite (http://mesquiteproject.org/mesquite/mesquite.html) is particularly
well suited for examining the evolution of characters on phylogenies.
#### Existing libraries for tree manipulation and rendering trees
ape (http://ape.mpl.ird.fr) - R tools for manipulating and analyzing trees,
extensive functionality for rendering trees and showing data on the trees. It
is now very widely used for analyzing evolutionary data on phylogenies, and
increasingly often for rendering trees for publication. There are quite a few
other R packages for examining phylogenies, these are summarize by Brian
O'Meara at http://cran.r-project.org/web/views/Phylogenetics.html.
Dendropy (http://packages.python.org/DendroPy/tutorial/index.html) - python
tools for manipulating and analyzing trees, not much for displaying them
#### Existing web-based viewers
svg's of trees. It produces very nice graphics, and
jstree (http://lh3lh3.users.sourceforge.net/jstree.shtml) - an editor for
onezoom (http://www.onezoom.org) - a set of static trees that can be viewed in
## The ideal phylogeny viewer
The ideal tool would be:
- Scalable, working well for trees that have a handful of tips up to millions
of tips (there are about 2 million described species, and probably at least 10
million currently living on the planet)
- Interactive, enabling the user to explore the tree (traversing different
parts, changing the zoom), manipulate the tree layout (move tips around,
control node density, rotate subtrees, etc), subset the tree according to data
(such as removing all nodes for species that were described after a particular
date), and control what data is shown about nodes and edges (toggle the names,
control what color the nodes are)
- Have meaningful transitions when different portions of the tree are shown
(eg, dynamic layout that optimizes the view of the tree as nodes are added or
removed) and data are shown or hidden
## Example use cases (not yet implemented)
### Show the history of biological exploration
Show a tree with all described species and the date they were first described.
Place a slider below the tree that goes from the first described species to the
present day. As the user moves the slider, only the species that were described
before the indicated date are shown. This allows the user to see how some
groups filled in slowly and then quickly in a burst of discovery, and how
entire new groups were discovered and then expanded.
- When the slider is all the way to the left there would be <50 species, when
to the right there would be hundreds of thousands or millions. The layout would
dynamically change to accommodate this change in density.
- As the slider moves, it would be very cool to change the color or
transparency of the nodes and edges that were just added in or are about to be
removed. That would make it easy to tell at a glance what is changing without
depending on the motion itself.
The simplest way to implement this would be to parse the taxonomy of itis.gov
and use it as a proxy for the tree of life. itis is a database of categories,
not relationships, but we don't have a single tree with true relationships yet.
itis also has the dates that species were described on, which is convenient.
itis_sql_to_json.py parses the tree structure and date. It propagates dates to
internal nodes, so that given year Y on the slider all nodes with a date great
than Y would be removed from the current view.
I used itis_sql_to_json.py to generate siphonophorae.json, which can be used as
a test dataset to get the viewer working. We can then create larger and larger
subtrees, up to all the species in itis.
### A widget for embedding interactive phylogenies into web sites.
The owner of the site could hard-code the tree into the web page. The widget
could also be configured to allow users to paste trees into the browser. In
both cases, newick encoded strings would be accepted.