Polytomies and tree comparison

Question

[Lee Katz]

Hi, I want to compare trees using some standard metrics. However, the metrics give crazy values when they come across polytomies. They give heavy penalties for small differences in the polytomy. The online literature usually advises you to split them randomly or to split them in the order that they appear in the tree, but again, this causes small changes that cause heavy penalties in the metrics.

major edit: After talking with a colleague, the problem can be simplified to collapsing low-confidence ancestor nodes and sorting polytomies. Does anyone have any code to do that in either R or Perl? (or other!?)

Answer 1

[jhc]

ETE has an experimental option to expand polytomies into all possible binary trees during tree comparison (i.e. Robinson-Foulds). The best score after comparing all combinations of binary trees is reported.

However, the method will explote with more than a few multifurcations or large topologies, so only useful for small trees.

This feature is not documented yet, but is part of the latest version. Check the expand_polytomies flag in the following functions:

• http://etetoolkit.org/docs/latest/reference/reference_tree.html#ete3.TreeNode.compare
• http://etetoolkit.org/docs/latest/reference/reference_tree.html#ete3.TreeNode.expand_polytomies
• http://etetoolkit.org/docs/latest/reference/reference_tree.html#ete3.TreeNode.robinson_foulds