Visualizing the Tree of Life

Classifying Tree Visualizations

Unlike the Tree of Life itself, there is no ontological reason to assume that the realm of known visualizations for tree-structured data should be representable with a single branching taxonomy. A possible faceted classification is presented below. A faceted classification provides a controled vocabulary for describing the visualizations. The seven facets include Structure substance, Interactivity, Spatial Description, Displayed Data, Data Access, View, and Navigation.

It would be worth considering which types of visualizations and characteristics are appropriate for which of our audiences. Interestingly, although I would have expected that the classificaiton could provide some guidelines as to which types of visualizations would do better on a heuristic evaluation, it seems to have limited usefulness for this.

Preliminary taxonomy of tree representations

revised Jan 05

Structure Substance
With what substance is the structural information visualized?

• characters (a matrix)
• indented text
• branches - a classic "tree"
• shapes
  • nesting (treeMap)
  • proximity (Prometheus)

Interactivity afforded by the visualization

• static (published diagrams, see "Assembling the Tree of Life")
• interactive
  • custom-prepared content (eg. Dennis Wall's Phylomongraphy)
  • dynamically-generated from a database
    • the data could change, but doesn't in the example. (hyperbolic tree, BIOT)
    • the data is actually constantly changing (eg. GenBank)

Spatial Description of the visualization

• Dimensionality
  • linear (eg. text, many static branching diagrams, Prometheus)
  • 2-D (treeMap, Understanding Evolution)
  • Hyperbolic (PEG, Green Plant Branch)
  • 3-D (eg. beamtrees, WALRUS)
• Orientation
  Many people will unconsciously read meaning into the left-right or bottom-top progression of diagrams, even where biologists intend none.
  • left - right (Paup, many published trees)
  • bottom - top (MacClade)
  • center-out (Understanding Evolution, WALRUS)
  • space-filling (basic treeMap, because location isn't meaningful, only nesting)

Displayed Data Layers
In addition to the basic structure and labeled leaf nodes, how many layers of information are conveyed by the visualization? Examples include geographical location (Dennis Wall's Phylomongraphy, future MapSteadi) or visualizing the distribution of a character state, perhaps the presence of a one-way gut or a more obscure synapomorphy (MacClade). This facet provides a list of common types of data included on tree visualizations. Any visualization may use none, one or more of these descriptors, and new visualizations may require additional descriptors. Interactive visualizations may provide some of this data on mouse-over (see SmartMoney.com's Map of the Market, a treeMap), blurring the line between data display and data access.

• Labeling nodes
  Note when all nodes are treated the same, vs. when leaf nodes are labeled differently than other nodes.
  • Clade names
  • names for other groupings
  • pictograms of representative organisms
  • statistics or other numeric properties of nodes
• Branch Lengths - branch lengths may specifically reflect more information than just the branching structure.
  • Distance or Time
  • ambiguous
  • Geologic Time (see chordates diagram from "Assembling the Tree of Life")
  • Molecular
• Branch properties - any information encoded in the color or width of branches or rectangles , or indicated with percentages or symbols
  • Support
  • synapomorphies or distribution of a character state (MacClade, diagram from "Assembling the Tree of Life" )
  • number of samples/species
• Navigation cues (Degree of Interest tree browser)

Data Access
Does visualization intend to provide access to additional data (dichotomous keys, taxonTree, Dennis Wall's Phylomongraphy)? This facet provides a list of common types of data accessed through tree visualizations. Any visualization may use none, one or more of these descriptors, and new visualizations may require additional descriptors.

• Access to data within the same resource.
  Interactive applications might provide a coordinated viewing window, or multiple windows (PEG)
  • descriptive text (dichotomous key in field guide, chordates diagram from "Assembling the Tree of Life")
  • geography ( Dennis Wall's Phylomongraphy, future MapSteadi)
  • bibliography (BIOT)
  • search results from a database
• Reference to information in another source
  • hyperlinks ( taxonTree, Green Plant Branch)

View(s) on tree-structure
A visualization will have at least one view extent and perspective, and possibly more.

• Extent
• Entire - all-at-once (most static diagrams)
• Focus - a local view (one page of tol website )
• Focus-plus-context (hyperbolic tree, Degree of Interest tree browser)
• Summary - an index to particular parts of the tree ( tol website 's Popular Pages feature, Digimorph.org)
• Perspective
• depth first, following a lineage (lists like this or tol website as a whole )
• breadth first, cross-section, (many dichotomous keys, treeMap)
• combination
• path-plus-breadth lineage browser (BIOT, GenBank, old parts of tol website)
• flexible, user-specified (taxonTree)

Navigation
When Focus views, multiple views or 3-D environments require that the viewer be able to interact with the visualization, what types of movements are provided?

• 3-D VR drag-to-turn-object style manipulation
• a focus window moving across the structure
• Contiguous - Move window locally only.
• zoom on breadth,
• local lineage-following on depth
• Jump the window around
• by access via another visualization
• by an Index or Summary structure ( Popular Pages display on the tol website)
• by Search (GenBank, Google's Directory)

Additional Interesting Categories

• summarization of a tree
• visualization of multiple trees and their relationships to each other
  • comparisons (DoubleTree, TreeJuxtaposer, Prometheus)
  • tree space (TreeSetViz )