Visualization tool for PLATO

Beta version

 ©Kent Vander Velden and Gavin Naylor

Iowa State University


Email:

graphix@iastate.edu

Address:
Dirac Science Library,
Florida State University,
Tallahassee, FL 32306-4120

USA
Tel:
850-644-3700

Fax:
850-644-9829



Visualization tool for PLATO
Grassly and Rambaut (1997) developed software (PLATO) to identify sub-regions of mutiple alignments whose likelihood scores suggested a phylogenetic signal different from that yielded by the overall analysis. We have augmented Grassly and Rambaut's approach with a visualization tool that allows the user to visually inspect which regions of the alignment have significantly low likelihoods (relative to a null distribution generated using Monte Carlo simulation).

Motivation:

Phylogenetic trees are estimated from DNA sequences by searching for the tree that best explains the data given a model of evolutionary change. The tree selected as optimal is the one providing the best collective fit over the entire sequence. However, it is possible that local patches of sequence exist for which the overall best tree is not locally optimal. Such "patches" of sequence might reflect regions of DNA subject to functional constraints not accommodated by the model of evolutionary change.Alternatively, they might reflect regions that have been subjected to recombination or horizontal transfer events.


Principle:

Platoís procedure is carried out over window (5 bases to half of the sequence alignment). Likelihood scores that are statistically different from null expectation are computed for each site in the alignment, over all window sizes analyzed, and are presented as a movie in which each frame corresponds to the output for a particular window size.


Features:

Results of the simulation are presented projected on to the 3-D structure of cytochrome B (the molecule over which the moving window analysis was applied) . The visualization allows us to quickly gain a sense of the spatial scales at which signal heterogeneity due to forces not accommodated by the evolutionary model become apparent. Patches that have significantly low likelihoods at small window sizes tend to be swamped out when larger window sizes are used. Large regions of significantly low likelihood, not identified at smaller window sizes, often become apparent when windows of a larger size are used. This is likely because many of the subtle tendencies expressed over large scales are not picked up at small window sizes.


Demo:

Movies for the two previously described features are available. (1) Plato vs. Rate of change (2) Plato vs. Tree
You will probably need one of the following media plug ins (freeware):

Real Player Quick Time Media Player

Features:

Results of the simulation are presented projected on to the 3-D structure of cytochrome B (the molecule over which the moving window analysis was applied) . The visualization allows us to quickly gain a sense of the spatial scales at which signal heterogeneity due to forces not accommodated by the evolutionary model become apparent.
Patches that have significantly low likelihoods at small window sizes tend to be swamped out when larger window sizes are used. Large regions of significantly low likelihood, not identified at smaller window sizes, often become apparent when windows of a larger size are used. This is likely because many of the subtle tendencies expressed over large scales are not picked up at small window sizes.


 

Click to enlarge
Heterogeneity and rate of change:

On the left panel, the regions identified with significantly low likelihoods change (blue to red gradient).
On the right panel, we can see that these regions identified are roughly correlated with highly variable regions.
The bottom bars show the "PLATO" information on linear sequence. 

Heterogeneity and tree topology:

This feature shows the effect of the inconsistent regions (left panel) on the tree topology (right panel). The regions are deleted and the tree infered is compared to the assumed "true" tree (percentage).


Availability:

A Unix version of the software is available.

 

Naylor Laboratory
Copyright © 2001, Iowa State University, all rights reserved.