Working With Tree Data Structures

Trees

Trees are a widely-used type of data structure that emulates a tree design with a set of linked nodes. Formally, a tree is considered an acyclic and connected graph. Each node in a tree has zero or more child nodes, which are below it in the tree (by convention, trees grow down, not up as they do in nature). A node that has a child is called the child’s parent node (or ancestor node, or superior). A node has at most one parent.

The height of a node is the length of the longest downward path to a leaf from that node. The height of the root is the height of the tree. The depth of a node is the length of the path to its root (i.e., its root path).

  • The topmost node in a tree is called the root node. Being the topmost node, the root node will not have parents. It is the node at which operations on the tree commonly begin (although some algorithms begin with the leaf nodes and work up ending at the root). All other nodes can be reached from it by following edges or links. Every node in a tree can be seen as the root node of the subtree rooted at that node.
  • Nodes at the bottommost level of the tree are called leaf nodes. Since they are at the bottommost level, they do not have any children.
  • An internal node or inner node is any node of a tree that has child nodes and is thus not a leaf node.
  • A subtree is a portion of a tree data structure that can be viewed as a complete tree in itself. Any node in a tree T, together with all the nodes below it, comprise a subtree of T. The subtree corresponding to the root node is the entire tree; the subtree corresponding to any other node is called a proper subtree (in analogy to the term proper subset).

In bioinformatics, trees are the result of many analyses, such as phylogenetics or clustering. Although each case entails specific considerations, many properties remains constant among them. In this respect, ETE is a python toolkit that assists in the automated manipulation, analysis and visualization of any type of hierarchical trees. It provides general methods to handle and visualize tree topologies, as well as specific modules to deal with phylogenetic and clustering trees.

Reading and Writing Newick Trees

The Newick format is one of the most widely used standard representation of trees in bioinformatics. It uses nested parentheses to represent hierarchical data structures as text strings. The original newick standard is able to encode information about the tree topology, branch distances and node names. Nevertheless, it is not uncommon to find slightly different formats using the newick standard.

ETE can read and write many of them:

FORMAT DESCRIPTION SAMPLE
0 flexible with support values ((D:0.723274,F:0.567784)1.000000:0.067192,(B:0.279326,H:0.756049)1.000000:0.807788);
1 flexible with internal node names ((D:0.723274,F:0.567784)E:0.067192,(B:0.279326,H:0.756049)B:0.807788);
2 all branches + leaf names + internal supports ((D:0.723274,F:0.567784)1.000000:0.067192,(B:0.279326,H:0.756049)1.000000:0.807788);
3 all branches + all names ((D:0.723274,F:0.567784)E:0.067192,(B:0.279326,H:0.756049)B:0.807788);
4 leaf branches + leaf names ((D:0.723274,F:0.567784),(B:0.279326,H:0.756049));
5 internal and leaf branches + leaf names ((D:0.723274,F:0.567784):0.067192,(B:0.279326,H:0.756049):0.807788);
6 internal branches + leaf names ((D,F):0.067192,(B,H):0.807788);
7 leaf branches + all names ((D:0.723274,F:0.567784)E,(B:0.279326,H:0.756049)B);
8 all names ((D,F)E,(B,H)B);
9 leaf names ((D,F),(B,H));
100 topology only ((,),(,));

Formats labeled as flexible allow for missing information. For instance, format 0 will be able to load a newick tree even if it does not contain branch support information (it will be initialized with the default value). However, format 2 would raise an exception. In other words, if you want to control that your newick files strictly follow a given pattern you should use strict format definitions.

Reading newick trees

In order to load a tree from a newick text string you can use the constructor TreeNode or its Tree alias, provided by the main module ete2. You will only need to pass a text string containing the newick structure and the format that should be used to parse it (0 by default). Alternatively, you can pass the path to a text file containing the newick string.

from ete2 import Tree

# Loads a tree structure from a newick string. The returned variable ’t’ is the root node for the tree.
t = Tree("(A:1,(B:1,(E:1,D:1):0.5):0.5);" )

# Load a tree structure from a newick file.
t = Tree("genes_tree.nh")

# You can also specify the newick format. For instance, for named internal nodes we will use format 1.
t = Tree("(A:1,(B:1,(E:1,D:1)Internal_1:0.5)Internal_2:0.5)Root;", format=1)

Writing newick trees

Any ETE tree instance can be exported using newick notation using the Tree.write() method, which is available in any tree node instance. It also allows for format selection (Reading and Writing Newick Trees), so you can use the same function to convert between newick formats.

from ete2 import Tree

# Loads a tree with internal node names
t = Tree("(A:1,(B:1,(E:1,D:1)Internal_1:0.5)Internal_2:0.5)Root;", format=1)

# And prints its newick using the default format

print t.write() # (A:1.000000,(B:1.000000,(E:1.000000,D:1.000000)1.000000:0.500000)1.000000:0.500000);

# To print the internal node names you need to change the format:

print t.write(format=1) # (A:1.000000,(B:1.000000,(E:1.000000,D:1.000000)Internal_1:0.500000)Internal_2:0.500000);

# We can also write into a file
t.write(format=1, outfile="new_tree.nw")

Understanding ETE Trees

Any tree topology can be represented as a succession of nodes connected in a hierarchical way. Thus, for practical reasons, ETE makes no distinction between tree and node concepts, as any tree can be represented by its root node. This allows to use any internal node within a tree as another sub-tree instance.

Once trees are loaded, they can be manipulated as normal python objects. Given that a tree is actually a collection of nodes connected in a hierarchical way, what you usually see as a tree will be the root node instance from which the tree structure is hanging. However, every node within a ETE’s tree structure can be also considered a subtree. This means, for example, that all the operational methods that we will review in the following sections are available at any possible level within a tree. Moreover, this feature will allow you to separate large trees into smaller partitions, or concatenate several trees into a single structure. For this reason, you will find that the TreeNode and Tree classes are synonymous.

Basic tree attributes

Each tree node has two basic attributes used to establish its position in the tree: TreeNode.up and TreeNode.children. The first is a pointer to parent’s node, while the later is a list of children nodes. Although it is possible to modify the structure of a tree by changing these attributes, it is strongly recommend not to do it. Several methods are provided to manipulate each node’s connections in a safe way (see Comparing Trees).

In addition, three other basic attributes are always present in any tree node instance:

Method Description Default value
TreeNode.dist stores the distance from the node to its parent (branch length). Default value = 1.0 1.0
TreeNode.support informs about the reliability of the partition defined by the node (i.e. bootstrap support) 1.0
TreeNode.name Custom node’s name. NoName

In addition, several methods are provided to perform basic operations on tree node instances:

Method Description
TreeNode.is_leaf() returns True if node has no children
TreeNode.is_root() returns True if node has no parent
TreeNode.get_tree_root() returns the top-most node within the same tree structure as node
len(TreeNode) returns the number of leaves under node
print node prints a text-based representation of the tree topology under node
if node in tree returns true if node is a leaf under tree
for leaf in node iterates over all leaves under node
TreeNode.show() Explore node graphically using a GUI.

This is an example on how to access such attributes:

from ete2 import Tree
t = Tree()
# We create a random tree topology
t.populate(15)
print t
print t.children
print t.get_children()
print t.up
print t.name
print t.dist
print t.is_leaf()
print t.get_tree_root()
print t.children[0].get_tree_root()
print t.children[0].children[0].get_tree_root()
# You can also iterate over tree leaves using a simple syntax
for leaf in t:
  print leaf.name

Root node on unrooted trees?

When a tree is loaded from external sources, a pointer to the top-most node is returned. This is called the tree root, and it will exist even if the tree is conceptually considered as unrooted. This is, the root node can be considered as the master node, since it represents the whole tree structure. Unrooted trees can be identified as trees in which master root node has more than two children.

from ete2 import Tree
unrooted_tree = Tree( "(A,B,(C,D));" )
print unrooted_tree
#
#     /-A
#    |
#----|--B
#    |
#    |     /-C
#     \---|
#          \-D

rooted_tree = Tree( "((A,B).(C,D));" )
print rooted_tree
#
#          /-A
#     /---|
#    |     \-B
#----|
#    |     /-C
#     \---|
#          \-D

Browsing trees (traversing)

One of the most basic operations for tree analysis is tree browsing. This is, essentially, visiting nodes within a tree. ETE provides a number of methods to search for specific nodes or to navigate over the hierarchical structure of a tree.

Getting Leaves, Descendants and Node’s Relatives

TreeNode instances contain several functions to access their descendants. Available methods are self explanatory:

Traversing (browsing) trees

Often, when processing trees, all nodes need to be visited. This is called tree traversing. There are different ways to traverse a tree structure depending on the order in which children nodes are visited. ETE implements the three most common strategies: preorder, levelorder and postorder. The following scheme shows the differences in the strategy for visiting nodes (note that in both cases the whole tree is browsed):

  • preorder: 1)Visit the root, 2) Traverse the left subtree , 3) Traverse the right subtree.
  • postorder: 1) Traverse the left subtree , 2) Traverse the right subtree, 3) Visit the root
  • levelorder (default): every node on a level before is visited going to a lower level

Note

  • Preorder traversal sequence: F, B, A, D, C, E, G, I, H (root, left, right)
  • Inorder traversal sequence: A, B, C, D, E, F, G, H, I (left, root, right); note how this produces a sorted sequence
  • Postorder traversal sequence: A, C, E, D, B, H, I, G, F (left, right, root)
  • Level-order traversal sequence: F, B, G, A, D, I, C, E, H

Every node in a tree includes a TreeNode.traverse() method, which can be used to visit, one by one, every node node under the current partition. In addition, the TreeNode.iter_descendants() method can be set to use either a post- or a preorder strategy. The only different between TreeNode.traverse() and TreeNode.iter_descendants() is that the first will include the root node in the iteration.

strategy can take one of the following values: "postorder", "preorder" or "levelorder"

# we load a tree
t = Tree('((((H,K)D,(F,I)G)B,E)A,((L,(N,Q)O)J,(P,S)M)C);', format=1)

for node in t.traverse("postorder"):
  # Do some analysis on node
  print node.name

# If we want to iterate over a tree excluding the root node, we can
# use the iter_descendant method
for node in t.iter_descendants("postorder"):
  # Do some analysis on node
  print node.name

Additionally, you can implement your own traversing function using the structural attributes of nodes. In the following example, only nodes between a given leaf and the tree root are visited.

from ete2 import Tree
tree = Tree( "(A:1,(B:1,(C:1,D:1):0.5):0.5);" )

# Browse the tree from a specific leaf to the root
node = t.search_nodes(name="C")[0]
while node:
   print node
   node = node.up

Advanced traversing (stopping criteria)

Collapsing nodes while traversing (custom is_leaf definition)

From version 2.2, ETE supports the use of the is_leaf_fn argument in most of its traversing functions. The value of is_leaf_fn is expected to be a pointer to any python function that accepts a node instance as its first argument and returns a boolean value (True if node should be considered a leaf node).

By doing so, all traversing methods will use such a custom function to decide if a node is a leaf. This becomes specially useful when dynamic collapsing of nodes is needed, thus avoiding to prune the same tree in many different ways.

For instance, given a large tree structure, the following code will export the newick of the pruned version of the topology, where nodes grouping the same tip labels are collapsed.

from ete2 import Tree
def collapsed_leaf(node):
    if len(node2labels[node]) == 1:
       return True
    else:
       return False

t = Tree("((((a,a,a)a,a)aa, (b,b)b)ab, (c, (d,d)d)cd);", format=1)
print t
# We create a cache with every node content
node2labels = t.get_cached_content(store_attr="name")
print t.write(is_leaf_fn=collapsed_leaf)
#             /-a
#            |
#          /-|--a
#         |  |
#       /-|   \-a
#      |  |
#    /-|   \-a
#   |  |
#   |  |   /-b
# --|   \-|
#   |      \-b
#   |
#   |   /-c
#    \-|
#      |   /-d
#       \-|
#          \-d

# We can even load the collapsed version as a new tree
t2 = Tree( t.write(is_leaf_fn=collapsed_leaf) )
print t2
#       /-aa
#    /-|
#   |   \-b
# --|
#   |   /-c
#    \-|
#       \-d

Another interesting use of this approach is to find the first matching nodes in a given tree that match a custom set of criteria, without browsing the whole tree structure.

Let’s say we want get all deepest nodes in a tree whose branch length is larger than one:

from ete2 import Tree
t = Tree("(((a,b)ab:2, (c, d)cd:2)abcd:2, ((e, f):2, g)efg:2);", format=1)
def processable_node(node):
    if node.dist > 1:
       return True
    else:
       return False

for leaf in t.iter_leaves(is_leaf_fn=processable_node):
    print leaf

#       /-a
#    /-|
#   |   \-b
# --|
#   |   /-c
#    \-|
#       \-d
#
#       /-e
#    /-|
# --|   \-f
#   |
#    \-g

Iterating instead of Getting

As commented previously, methods starting with get_ are all prepared to return results as a closed list of items. This means, for instance, that if you want to process all tree leaves and you ask for them using the TreeNode.get_leaves() method, the whole tree structure will be browsed before returning the final list of terminal nodes. This is not a problem in most of the cases, but in large trees, you can speed up the browsing process by using iterators.

Most get_ methods have their homologous iterator functions. Thus, TreeNode.get_leaves() could be substituted by TreeNode.iter_leaves(). The same occurs with TreeNode.iter_descendants() and TreeNode.iter_search_nodes().

When iterators are used (note that is only applicable for looping), only one step is processed at a time. For instance, TreeNode.iter_search_nodes() will return one match in each iteration. In practice, this makes no differences in the final result, but it may increase the performance of loop functions (i.e. in case of finding a match which interrupts the loop).

Finding nodes by their attributes

Both terminal and internal nodes can be located by searching along the tree structure. Several methods are available:

method Description
t.search_nodes(attr=value) Returns a list of nodes in which attr is equal to value, i.e. name=A
t.iter_search_nodes(attr=value) Iterates over all matching nodes matching attr=value. Faster when you only need to get the first occurrence
t.get_leaves_by_name(name) Returns a list of leaf nodes matching a given name. Only leaves are browsed.
t.get_common_ancestor([node1, node2, node3]) Return the first internal node grouping node1, node2 and node3
t&”A” Shortcut for t.search_nodes(name=”A”)[0]

Search_all nodes matching a given criteria

A custom list of nodes matching a given name can be easily obtain through the TreeNode.search_node() function.

from ete2 import Tree
t = Tree( '((H:1,I:1):0.5, A:1, (B:1,(C:1,D:1):0.5):0.5);' )
print t
#                    /-H
#          /--------|
#         |          \-I
#         |
#---------|--A
#         |
#         |          /-B
#          \--------|
#                   |          /-C
#                    \--------|
#                              \-D

# I get D
D = t.search_nodes(name="D")[0]

# I get all nodes with distance=0.5
nodes = t.search_nodes(dist=0.5)
print len(nodes), "nodes have distance=0.5"

# We can limit the search to leaves and node names (faster method).
D = t.get_leaves_by_name(name="D")
print D

Search nodes matching a given criteria (iteration)

A limitation of the TreeNode.search_nodes() method is that you cannot use complex conditional statements to find specific nodes. When search criteria is too complex, you may need to create your own search function.

from ete2 import Tree

def search_by_size(node, size):
    "Finds nodes with a given number of leaves"
    matches = []
    for n in node.traverse():
       if len(n) == size:
          matches.append(n)
    return matches

t = Tree()
t.populate(40)
# returns nodes containing 6 leaves
search_by_size(t, size=6)

Find the first common ancestor

Searching for the first common ancestor of a given set of nodes it is a handy way of finding internal nodes.

from ete2 import Tree
t = Tree( "((H:0.3,I:0.1):0.5, A:1, (B:0.4,(C:0.5,(J:1.3, (F:1.2, D:0.1):0.5):0.5):0.5):0.5);" )
print t
ancestor = t.get_common_ancestor("C", "J", "B")

Custom searching functions

A limitation of the previous methods is that you cannot use complex conditional statements to find specific nodes. However you can user traversing methods to meet your custom filters. A possible general strategy would look like this:

from ete2 import Tree
t = Tree("((H:0.3,I:0.1):0.5, A:1, (B:0.4,(C:1,D:1):0.5):0.5);")
# Create a small function to filter your nodes
def conditional_function(node):
    if node.dist > 0.3:
        return True
    else:
        return False

# Use previous function to find matches. Note that we use the traverse
# method in the filter function. This will iterate over all nodes to
# assess if they meet our custom conditions and will return a list of
# matches.
matches = filter(conditional_function, t.traverse())
print len(matches), "nodes have ditance >0.3"

# depending on the complexity of your conditions you can do the same
# in just one line with the help of lambda functions:
matches = filter(lambda n: n.dist>0.3 and n.is_leaf(), t.traverse() )
print len(matches), "nodes have ditance >0.3 and are leaves"

Shortcuts

Finally, ETE implements a built-in method to find the first node matching a given name, which is one of the most common tasks needed for tree analysis. This can be done through the operator & (AND). Thus, TreeNode&”A” will always return the first node whose name is “A” and that is under the tree “MyTree”. The syntaxis may seem confusing, but it can be very useful in some situations.

from ete2 import Tree
t = Tree("((H:0.3,I:0.1):0.5, A:1, (B:0.4,(C:1,(J:1, (F:1, D:1):0.5):0.5):0.5):0.5);")
# Get the node D in a very simple way
D = t&"D"
# Get the path from B to the root
node = D
path = []
while node.up:
  path.append(node)
  node = node.up
print t
# I substract D node from the total number of visited nodes
print "There are", len(path)-1, "nodes between D and the root"
# Using parentheses you can use by-operand search syntax as a node
# instance itself
Dsparent= (t&"C").up
Bsparent= (t&"B").up
Jsparent= (t&"J").up
# I check if nodes belong to certain partitions
print "It is", Dsparent in Bsparent, "that C's parent is under B's ancestor"
print "It is", Dsparent in Jsparent, "that C's parent is under J's ancestor"

Checking the monophyly of attributes within a tree

Although monophyly is actually a phylogenetic concept used to refer to a set of species that group exclusively together within a tree partition, the idea can be easily exported to any type of trees.

Therefore, we could consider that a set of values for a given node attribute present in our tree is monophyletic, if such values group exclusively together as a single tree partition. If not, the corresponding relationship connecting such values (para or poly-phyletic) could be also be inferred.

The TreeNode.check_monophyly() method will do so when a given tree is queried for any custom attribute.

from ete2 import Tree
t =  Tree("((((((a, e), i), o),h), u), ((f, g), j));")
print t

#                   /-a
#                /-|
#             /-|   \-e
#            |  |
#          /-|   \-i
#         |  |
#       /-|   \-o
#      |  |
#    /-|   \-h
#   |  |
#   |   \-u
# --|
#   |      /-f
#   |   /-|
#    \-|   \-g
#      |
#       \-j


# We can check how, indeed, all vowels are not monophyletic in the
# previous tree, but polyphyletic (a foreign label breaks its monophyly)
print t.check_monophyly(values=["a", "e", "i", "o", "u"], target_attr="name")

# however, the following set of vowels are monophyletic
print t.check_monophyly(values=["a", "e", "i", "o"], target_attr="name")

# A special case of polyphyly, called paraphyly, is also used to
# define certain type of grouping. See this wikipedia article for
# disambiguation: http://en.wikipedia.org/wiki/Paraphyly
print t.check_monophyly(values=["i", "o"], target_attr="name")

Finally, the TreeNode.get_monophyletic() method is also provided, which allows to return a list of nodes within a tree where a given set of attribute values are monophyletic. Note that, although a set of values are not monophyletic regarding the whole tree, several independent monophyletic partitions could be found within the same topology.

For instance, in the following example, all clusters within the same tree exclusively grouping a custom set of annotations are obtained.

from ete2 import Tree
t =  Tree("((((((4, e), i), o),h), u), ((3, 4), (i, june)));")
# we annotate the tree using external data
colors = {"a":"red", "e":"green", "i":"yellow",
          "o":"black", "u":"purple", "4":"green",
          "3":"yellow", "1":"white", "5":"red",
          "june":"yellow"}
for leaf in t:
    leaf.add_features(color=colors.get(leaf.name, "none"))
print t.get_ascii(attributes=["name", "color"], show_internal=False)

#                   /-4, green
#                /-|
#             /-|   \-e, green
#            |  |
#          /-|   \-i, yellow
#         |  |
#       /-|   \-o, black
#      |  |
#    /-|   \-h, none
#   |  |
#   |   \-u, purple
# --|
#   |      /-3, yellow
#   |   /-|
#   |  |   \-4, green
#    \-|
#      |   /-i, yellow
#       \-|
#          \-june, yellow

print "Green-yellow clusters:"
# And obtain clusters exclusively green and yellow
for node in t.get_monophyletic(values=["green", "yellow"], target_attr="color"):
   print node.get_ascii(attributes=["color", "name"], show_internal=False)

# Green-yellow clusters:
#
#       /-green, 4
#    /-|
# --|   \-green, e
#   |
#    \-yellow, i
#
#       /-yellow, 3
#    /-|
#   |   \-green, 4
# --|
#   |   /-yellow, i
#    \-|
#       \-yellow, june

Note

When the target attribute is set to the “species” feature name, associated to any PhyloTree node, this method will accomplish with the standard phylogenetic definition of monophyly, polyphyly and paraphyly.

Caching tree content for faster lookup operations

If your program needs to access to the content of different nodes very frequently, traversing the tree to get the leaves of each node over and over will produce significant slowdowns in your algorithm. From version 2.2 ETE provides a convenient methods to cache frequent data.

The method TreeNode.get_cached_content() returns a dictionary in which keys are node instances and values represent the content of such nodes. By default, content is understood as a list of leave nodes, so looking up size or tip names under a given node will be instant. However, specific attributes can be cached by setting a custom store_attr value.

from ete2 import Tree
t = Tree()
t.populate(50)

node2leaves = t.get_cached_content()

# lets now print the size of each node without the need of
# recursively traverse
for n in t.traverse():
    print "node %s contains %s tips" %(n.name, len(node2leaves[n]))

Node annotation

Every node contains three basic attributes: name (TreeNode.name), branch length (TreeNode.dist) and branch support (TreeNode.support). These three values are encoded in the newick format. However, any extra data could be linked to trees. This is called tree annotation.

The TreeNode.add_feature() and TreeNode.add_features() methods allow to add extra attributes (features) to any node. The first allows to add one one feature at a time, while the second can be used to add many features with the same call.

Once extra features are added, you can access their values at any time during the analysis of a tree. To do so, you only need to access to the TreeNode.feature_name attributes.

Similarly, TreeNode.del_feature() can be used to delete an attribute.

import random
from ete2 import Tree
# Creates a tree
t = Tree( '((H:0.3,I:0.1):0.5, A:1, (B:0.4,(C:0.5,(J:1.3, (F:1.2, D:0.1):0.5):0.5):0.5):0.5);' )

# Let's locate some nodes using the get common ancestor method
ancestor=t.get_common_ancestor("J", "F", "C")
# the search_nodes method (I take only the first match )
A = t.search_nodes(name="A")[0]
# and using the shorcut to finding nodes by name
C= t&"C"
H= t&"H"
I= t&"I"

# Let's now add some custom features to our nodes. add_features can be
# used to add many features at the same time.
C.add_features(vowel=False, confidence=1.0)
A.add_features(vowel=True, confidence=0.5)
ancestor.add_features(nodetype="internal")

# Or, using the oneliner notation
(t&"H").add_features(vowel=False, confidence=0.2)

# But we can automatize this. (note that i will overwrite the previous
# values)
for leaf in t.traverse():
   if leaf.name in "AEIOU":
      leaf.add_features(vowel=True, confidence=random.random())
   else:
      leaf.add_features(vowel=False, confidence=random.random())

# Now we use these information to analyze the tree.
print "This tree has", len(t.search_nodes(vowel=True)), "vowel nodes"
print "Which are", [leaf.name for leaf in t.iter_leaves() if leaf.vowel==True]

# But features may refer to any kind of data, not only simple
# values. For example, we can calculate some values and store them
# within nodes.
#
# Let's detect leaf nodes under "ancestor" with distance higher thatn
# 1. Note that I'm traversing a subtree which starts from "ancestor"
matches = [leaf for leaf in ancestor.traverse() if leaf.dist>1.0]

# And save this pre-computed information into the ancestor node
ancestor.add_feature("long_branch_nodes", matches)

# Prints the precomputed nodes
print "These are nodes under ancestor with long branches", \
   [n.name for n in ancestor.long_branch_nodes]

# We can also use the add_feature() method to dynamically add new features.
label = raw_input("custom label:")
value = raw_input("custom label value:")
ancestor.add_feature(label, value)
print "Ancestor has now the [", label, "] attribute with value [", value, "]"

Unfortunately, newick format does not support adding extra features to a tree. Because of this drawback, several improved formats haven been (or are being) developed to read and write tree based information. Some of these new formats are based in a completely new standard (Phylogenetic XML standards), while others are extensions of the original newick formar (NHX http://phylosoft.org/NHX/http://phylosoft.org/NHX/).

Currently, ETE includes support for the New Hampshire eXtended format (NHX), which uses the original newick standard and adds the possibility of saving additional date related to each tree node. Here is an example of a extended newick representation in which extra information is added to an internal node:

(A:0.35,(B:0.72,(D:0.60,G:0.12):0.64[&&NHX:conf=0.01:name=INTERNAL]):0.56);

As you can notice, extra node features in the NHX format are enclosed between brackets. ETE is able to read and write features using such format, however, the encoded information is expected to be exportable as plain text.

The NHX format is automatically detected when reading a newick file, and the detected node features are added using the TreeNode.add_feature() method. Consequently, you can access the information by using the normal ETE’s feature notation: node.feature_name. Similarly, features added to a tree can be included within the normal newick representation using the NHX notation. For this, you can call the TreeNode.write() method using the features argument, which is expected to be a list with the features names that you want to include in the newick string. Note that all nodes containing the suplied features will be exposed into the newick string. Use an empty features list (features=[]) to include all node’s data into the newick string.

import random
from ete2 import Tree
# Creates a normal tree
t = Tree('((H:0.3,I:0.1):0.5, A:1,(B:0.4,(C:0.5,(J:1.3,(F:1.2, D:0.1):0.5):0.5):0.5):0.5);')
print t
# Let's locate some nodes using the get common ancestor method
ancestor=t.get_common_ancestor("J", "F", "C")
# Let's label leaf nodes
for leaf in t.traverse():
    if leaf.name in "AEIOU":
      leaf.add_features(vowel=True, confidence=random.random())
    else:
      leaf.add_features(vowel=False, confidence=random.random())

# Let's detect leaf nodes under "ancestor" with distance higher thatn
# 1. Note that I'm traversing a subtree which starts from "ancestor"
matches = [leaf for leaf in ancestor.traverse() if leaf.dist>1.0]

# And save this pre-computed information into the ancestor node
ancestor.add_feature("long_branch_nodes", matches)
print
print "NHX notation including vowel and confidence attributes"
print
print t.write(features=["vowel", "confidence"])
print
print "NHX notation including all node's data"
print

# Note that when all features are requested, only those with values
# equal to text-strings or numbers are considered. "long_branch_nodes"
# is not included into the newick string.
print t.write(features=[])
print
print "basic newick formats are still available"
print
print t.write(format=9, features=["vowel"])
# You don't need to do anything speciall to read NHX notation. Just
# specify the newick format and the NHX tags will be automatically
# detected.
nw = """
(((ADH2:0.1[&&NHX:S=human:E=1.1.1.1], ADH1:0.11[&&NHX:S=human:E=1.1.1.1])
:0.05[&&NHX:S=Primates:E=1.1.1.1:D=Y:B=100], ADHY:0.1[&&NHX:S=nematode:
E=1.1.1.1],ADHX:0.12[&&NHX:S=insect:E=1.1.1.1]):0.1[&&NHX:S=Metazoa:
E=1.1.1.1:D=N], (ADH4:0.09[&&NHX:S=yeast:E=1.1.1.1],ADH3:0.13[&&NHX:S=yeast:
E=1.1.1.1], ADH2:0.12[&&NHX:S=yeast:E=1.1.1.1],ADH1:0.11[&&NHX:S=yeast:E=1.1.1.1]):0.1
[&&NHX:S=Fungi])[&&NHX:E=1.1.1.1:D=N];
"""
# Loads the NHX example found at http://www.phylosoft.org/NHX/
t = Tree(nw)
# And access node's attributes.
for n in t.traverse():
    if hasattr(n,"S"):
       print n.name, n.S

Comparing Trees

Calculate distances between trees

The :Tree:`compare` function allows to calculate distances between two trees based on any node feature (i.e. name, species, other tags) using robinson-foulds and edge compatibility distances. It automatically handles differences in tree sizes, shared nodes and duplicated feature names.

  • result[“rf”] = robinson-foulds distance between the two trees. (average of robinson-foulds distances if target tree contained duplication and was split in several subtrees)
  • result[“max_rf”] = Maximum robinson-foulds distance expected for this comparison
  • result[“norm_rf”] = normalized robinson-foulds distance (from 0 to 1)
  • result[“effective_tree_size”] = the size of the compared trees, which are pruned to the common shared nodes.
  • result[“ref_edges_in_source”] = compatibility score of the target tree with respect to the source tree (how many edges in reference are found in the source)
  • result[“source_edges_in_ref”] = compatibility score of the source tree with respect to the reference tree (how many edges in source are found in the reference)
  • result[“source_subtrees”] = number of subtrees in the source tree (1 if do not contain duplications)
  • result[“common_edges”] = a set of common edges between source tree and reference
  • result[“source_edges”] = the set of edges found in the source tree
  • result[“ref_edges”] = the set of edges found in the reference tree
  • result[“treeko_dist”] = TreeKO speciation distance for comparisons including duplication nodes.

Robinson-foulds distance

Two tree topologies can be compared using ETE and the Robinson-Foulds (RF) metric. The method TreeNode.robinson_foulds() available for any ETE tree node allows to:

  • compare two tree topologies by their name labels (default) or any other annotated feature in the tree.
  • compare topologies of different size and content. When two trees contain a different set of labels, only shared leaves will be used.
  • examine size and content of matching and missing partitions. Since the method return the list of partitions found in both trees, details about matching partitions can be obtained easily.

In the following example, several of above mentioned features are shown:

from ete2 import Tree
t1 = Tree('(((a,b),c), ((e, f), g));')
t2 = Tree('(((a,c),b), ((e, f), g));')
rf, max_rf, common_leaves, parts_t1, parts_t2 = t1.robinson_foulds(t2)
print t1, t2
print "RF distance is %s over a total of %s" %(rf, max_rf)
print "Partitions in tree2 that were not found in tree1:", parts_t1 - parts_t2
print "Partitions in tree1 that were not found in tree2:", parts_t2 - parts_t1

# We can also compare trees sharing only part of their labels

t1 = Tree('(((a,b),c), ((e, f), g));')
t2 = Tree('(((a,c),b), (g, H));')
rf, max_rf, common_leaves, parts_t1, parts_t2 = t1.robinson_foulds(t2)

print t1, t2
print "Same distance holds even for partially overlapping trees"
print "RF distance is %s over a total of %s" %(rf, max_rf)
print "Partitions in tree2 that were not found in tree1:", parts_t1 - parts_t2
print "Partitions in tree1 that were not found in tree2:", parts_t2 - parts_t1

Modifying Tree Topology

Creating Trees from Scratch

If no arguments are passed to the TreeNode class constructor, an empty tree node will be returned. Such an orphan node can be used to populate a tree from scratch. For this, the TreeNode.up, and TreeNode.children attributes should never be used (unless it is strictly necessary). Instead, several methods exist to manipulate the topology of a tree:

from ete2 import Tree
t = Tree() # Creates an empty tree
A = t.add_child(name="A") # Adds a new child to the current tree root
                           # and returns it
B = t.add_child(name="B") # Adds a second child to the current tree
                           # root and returns it
C = A.add_child(name="C") # Adds a new child to one of the branches
D = C.add_sister(name="D") # Adds a second child to same branch as
                             # before, but using a sister as the starting
                             # point
R = A.add_child(name="R") # Adds a third child to the
                           # branch. Multifurcations are supported
# Next, I add 6 random leaves to the R branch names_library is an
# optional argument. If no names are provided, they will be generated
# randomly.
R.populate(6, names_library=["r1","r2","r3","r4","r5","r6"])
# Prints the tree topology
print t
#                     /-C
#                    |
#                    |--D
#                    |
#           /--------|                              /-r4
#          |         |                    /--------|
#          |         |          /--------|          \-r3
#          |         |         |         |
#          |         |         |          \-r5
#          |          \--------|
# ---------|                   |                    /-r6
#          |                   |          /--------|
#          |                    \--------|          \-r2
#          |                             |
#          |                              \-r1
#          |
#           \-B
# a common use of the populate method is to quickly create example
# trees from scratch. Here we create a random tree with 100 leaves.
t = Tree()
t.populate(100)

Deleting (eliminating) and Removing (detaching) nodes

As currently implemented, there is a difference between detaching and deleting a node. The former disconnects a complete partition from the tree structure, so all its descendants are also disconnected from the tree. There are two methods to perform this action: TreeNode.remove_child() and TreeNode.detach(). In contrast, deleting a node means eliminating such node without affecting its descendants. Children from the deleted node are automatically connected to the next possible parent. This is better understood with the following example:

from ete2 import Tree
# Loads a tree. Note that we use format 1 to read internal node names
t = Tree('((((H,K)D,(F,I)G)B,E)A,((L,(N,Q)O)J,(P,S)M)C);', format=1)
print "original tree looks like this:"
# This is an alternative way of using "print t". Thus we have a bit
# more of control on how tree is printed. Here i print the tree
# showing internal node names
print t.get_ascii(show_internal=True)
#
#                                        /-H
#                              /D-------|
#                             |          \-K
#                    /B-------|
#                   |         |          /-F
#          /A-------|          \G-------|
#         |         |                    \-I
#         |         |
#         |          \-E
#-NoName--|
#         |                    /-L
#         |          /J-------|
#         |         |         |          /-N
#         |         |          \O-------|
#          \C-------|                    \-Q
#                   |
#                   |          /-P
#                    \M-------|
#                              \-S
# Get pointers to specific nodes
G = t.search_nodes(name="G")[0]
J = t.search_nodes(name="J")[0]
C = t.search_nodes(name="C")[0]
# If we remove J from the tree, the whole partition under J node will
# be detached from the tree and it will be considered an independent
# tree. We can do the same thing using two approaches: J.detach() or
# C.remove_child(J)
removed_node = J.detach() # = C.remove_child(J)
# if we know print the original tree, we will see how J partition is
# no longer there.
print "Tree after REMOVING the node J"
print t.get_ascii(show_internal=True)
#                                        /-H
#                              /D-------|
#                             |          \-K
#                    /B-------|
#                   |         |          /-F
#          /A-------|          \G-------|
#         |         |                    \-I
#         |         |
#-NoName--|          \-E
#         |
#         |                    /-P
#          \C------- /M-------|
#                              \-S
# however, if we DELETE the node G, only G will be eliminated from the
# tree, and all its descendants will then hang from the next upper
# node.
G.delete()
print "Tree after DELETING the node G"
print t.get_ascii(show_internal=True)
#                                        /-H
#                              /D-------|
#                             |          \-K
#                    /B-------|
#                   |         |--F
#          /A-------|         |
#         |         |          \-I
#         |         |
#-NoName--|          \-E
#         |
#         |                    /-P
#          \C------- /M-------|
#                              \-S

Pruning trees

Pruning a tree means to obtain the topology that connects a certain group of items by removing the unnecessary edges. To facilitate this task, ETE implements the TreeNode.prune() method, which can be used by providing the list of terminal and/or internal nodes that must be kept in the tree.

From version 2.2, this function includes also the preserve_branch_length flag, which allows to remove nodes from a tree while keeping original distances among remaining nodes.

from ete2 import Tree
# Let's create simple tree
t = Tree('((((H,K),(F,I)G),E),((L,(N,Q)O),(P,S)));')
print "Original tree looks like this:"
print t
#
#                                        /-H
#                              /--------|
#                             |          \-K
#                    /--------|
#                   |         |          /-F
#          /--------|          \--------|
#         |         |                    \-I
#         |         |
#         |          \-E
#---------|
#         |                    /-L
#         |          /--------|
#         |         |         |          /-N
#         |         |          \--------|
#          \--------|                    \-Q
#                   |
#                   |          /-P
#                    \--------|
#                              \-S
# Prune the tree in order to keep only some leaf nodes.
t.prune(["H","F","E","Q", "P"])
print "Pruned tree"
print t
#
#                              /-F
#                    /--------|
#          /--------|          \-H
#         |         |
#---------|          \-E
#         |
#         |          /-Q
#          \--------|
#                    \-P
# Let's re-create the same tree again

Concatenating trees

Given that all tree nodes share the same basic properties, they can be connected freely. In fact, any node can add a whole subtree as a child, so we can actually cut & paste partitions. To do so, you only need to call the TreeNode.add_child() method using another tree node as a first argument. If such a node is the root node of a different tree, you will concatenate two structures. But caution!!, this kind of operations may result into circular tree structures if add an node’s ancestor as a new node’s child. Some basic checks are internally performed by the ETE topology related methods, however, a fully qualified check of this issue would affect seriously the performance of the method. For this reason, users themselves should take care about not creating circular structures by mistake.

from ete2 import Tree
# Loads 3 independent trees
t1 = Tree('(A,(B,C));')
t2 = Tree('((D,E), (F,G));')
t3 = Tree('(H, ((I,J), (K,L)));')
print "Tree1:", t1
#            /-A
#  ---------|
#           |          /-B
#            \--------|
#                      \-C
print "Tree2:", t2
#                      /-D
#            /--------|
#           |          \-E
#  ---------|
#           |          /-F
#            \--------|
#                      \-G
print "Tree3:", t3
#            /-H
#           |
#  ---------|                    /-I
#           |          /--------|
#           |         |          \-J
#            \--------|
#                     |          /-K
#                      \--------|
#                                \-L
# Locates a terminal node in the first tree
A = t1.search_nodes(name='A')[0]
# and adds the two other trees as children.
A.add_child(t2)
A.add_child(t3)
print "Resulting concatenated tree:", t1
#                                          /-D
#                                /--------|
#                               |          \-E
#                      /--------|
#                     |         |          /-F
#                     |          \--------|
#            /--------|                    \-G
#           |         |
#           |         |          /-H
#           |         |         |
#           |          \--------|                    /-I
#           |                   |          /--------|
#  ---------|                   |         |          \-J
#           |                    \--------|
#           |                             |          /-K
#           |                              \--------|
#           |                                        \-L
#           |
#           |          /-B
#            \--------|
#                      \-C

Copying (duplicating) trees

ETE provides several strategies to clone tree structures. The method TreeNode.copy() can be used to produce a new independent tree object with the exact topology and features as the original. However, as trees may involve many intricate levels of branches and nested features, 4 different methods are available to create a tree copy:

  • “newick”: Tree topology, node names, branch lengths and branch support values will be copied as represented in the newick string This method is based on newick format serialization works very fast even for large trees.
  • “newick-extended”: Tree topology and all node features will be copied based on the extended newick format representation. Only node features will be copied, thus excluding other node attributes. As this method is also based on newick serialisation, features will be converted into text strings when making the copy. Performance will depend on the tree size and the number and type of features being copied.
  • “cpickle”: This is the default method. The whole node structure and its content will be cloned based on the cPickle object serialization python approach. This method is slower, but recommended for full tree copying.
  • “deepcopy”: The whole node structure and its content is copied based on the standard “copy” Python functionality. This is the slowest method, but it allows to copy very complex objects even when attributes point to lambda functions.
from ete2 import Tree
t = Tree("((A, B)Internal_1:0.7, (C, D)Internal_2:0.5)root:1.3;", format=1)
# we add a custom annotation to the node named A
(t & "A").add_features(label="custom Value")
# we add a complex feature to the A node, consisting of a list of lists
(t & "A").add_features(complex=[[0,1], [2,3], [1,11], [1,0]])
print t.get_ascii(attributes=["name", "dist", "label", "complex"])

#                         /-A, 0.0, custom Value, [[0, 1], [2, 3], [1, 11], [1, 0]]
#          /Internal_1, 0.7
#         |               \-B, 0.0
# -root, 1.3
#         |               /-C, 0.0
#          \Internal_2, 0.5
#                         \-D, 0.0

# Newick copy will loose custom node annotations, complex features,
# but not names and branch values

print t.copy("newick").get_ascii(attributes=["name", "dist", "label", "complex"])

#                           /-A, 0.0
#            /Internal_1, 0.7
#           |               \-B, 0.0
# -NoName, 0.0
#           |               /-C, 0.0
#            \Internal_2, 0.5
#                           \-D, 0.0

# Extended newick copy will transfer custom annotations as text
# strings, so complex features are lost.

print t.copy("newick-extended").get_ascii(attributes=["name", "dist", "label", "complex"])

#                              /-A, 0.0, custom Value, __0_ 1__ _2_ 3__ _1_ 11__ _1_ 0__
#            /Internal_1, 0.7
#           |               \-B, 0.0
# -NoName, 0.0
#           |               /-C, 0.0
#            \Internal_2, 0.5
#                           \-D, 0.0

# The default pickle method will produce a exact clone of the
# original tree, where features are duplicated keeping their
# python data type.

print t.copy().get_ascii(attributes=["name", "dist", "label", "complex"])
print "first element in complex feature:", (t & "A").complex[0]

#                         /-A, 0.0, custom Value, [[0, 1], [2, 3], [1, 11], [1, 0]]
#          /Internal_1, 0.7
#         |               \-B, 0.0
# -root, 1.3
#         |               /-C, 0.0
#          \Internal_2, 0.5
#                         \-D, 0.0
# first element in complex feature: [0, 1]

Solving multifurcations

When a tree contains a polytomy (a node with more than 2 children), the method resolve_polytomy() can be used to convert the node into a randomly bifurcated structure in which branch lengths are set to 0. This is really not a solution for the polytomy but it allows to export the tree as a strictly bifurcated newick structure, which is a requirement for some external software.

The method can be used on a very specific node while keeping the rest of the tree intact by disabling the recursive flag.

from ete2 import Tree
t = Tree("(( (a, b, c), (d, e, f, g)), (f, i, h));")
print t

#             /-a
#            |
#         /--|--b
#        |   |
#        |    \-c
#     /--|
#    |   |    /-d
#    |   |   | y
#    |   |   |--e
#    |    \--|
# ---|       |--f
#    |       |
#    |        \-g
#    |
#    |    /-f
#    |   |
#     \--|--i
#        |
#         \-h


polynode = t.get_common_ancestor("a", "b")
polynode.resolve_polytomy(recursive=False)
print t

#                 /-b
#             /--|
#         /--|    \-c
#        |   |
#        |    \-a
#     /--|
#    |   |    /-d
#    |   |   |
#    |   |   |--e
#    |    \--|
# ---|       |--f
#    |       |
#    |        \-g
#    |
#    |    /-f
#    |   |
#     \--|--i
#        |
#         \-h


t.resolve_polytomy(recursive=True)
print t

#
#                 /-b
#             /--|
#         /--|    \-c
#        |   |
#        |    \-a
#        |
#     /--|            /-f
#    |   |        /--|
#    |   |    /--|    \-g
#    |   |   |   |
#    |    \--|    \-e
# ---|       |
#    |        \-d
#    |
#    |        /-i
#    |    /--|
#     \--|    \-h
#        |
#         \-f

Tree Rooting

Tree rooting is understood as the technique by with a given tree is conceptually polarized from more basal to more terminal nodes. In phylogenetics, for instance, this a crucial step prior to the interpretation of trees, since it will determine the evolutionary relationships among the species involved. The concept of rooted trees is different than just having a root node, which is always necessary to handle a tree data structure. Usually, the way in which a tree is differentiated between rooted and unrooted, is by counting the number of branches of the current root node. Thus, if the root node has more than two child branches, the tree is considered unrooted. By contrast, when only two main branches exist under the root node, the tree is considered rooted.

Having an unrooted tree means that any internal branch within the tree could be regarded as the root node, and there is no conceptual reason to place the root node where it is placed at the moment. Therefore, in an unrooted tree, there is no information about which internal nodes are more basal than others. By setting the root node between a given edge/branch of the tree structure the tree is polarized, meaning that the two branches under the root node are the most basal nodes. In practice, this is usually done by setting an outgroup node, which would represent one of these main root branches. The second one will be, obviously, the brother node. When you set an outgroup on unrooted trees, the multifurcations at the current root node are solved.

In order to root an unrooted tree or re-root a tree structure, ETE implements the TreeNode.set_outgroup() method, which is present in any tree node instance. Similarly, the TreeNode.unroot() method can be used to perform the opposite action.

from ete2 import Tree
# Load an unrooted tree. Note that three branches hang from the root
# node. This usually means that no information is available about
# which of nodes is more basal.
t = Tree('(A,(H,F)(B,(E,D)));')
print "Unrooted tree"
print t
#          /-A
#         |
#         |          /-H
#---------|---------|
#         |          \-F
#         |
#         |          /-B
#          \--------|
#                   |          /-E
#                    \--------|
#                              \-D
#
# Let's define that the ancestor of E and D as the tree outgroup.  Of
# course, the definition of an outgroup will depend on user criteria.
ancestor = t.get_common_ancestor("E","D")
t.set_outgroup(ancestor)
print "Tree rooteda at E and D's ancestor is more basal that the others."
print t
#
#                    /-B
#          /--------|
#         |         |          /-A
#         |          \--------|
#         |                   |          /-H
#---------|                    \--------|
#         |                              \-F
#         |
#         |          /-E
#          \--------|
#                    \-D
#
# Note that setting a different outgroup, a different interpretation
# of the tree is possible
t.set_outgroup( t&"A" )
print "Tree rooted at a terminal node"
print t
#                              /-H
#                    /--------|
#                   |          \-F
#          /--------|
#         |         |          /-B
#         |          \--------|
#---------|                   |          /-E
#         |                    \--------|
#         |                              \-D
#         |
#          \-A

Note that although rooting is usually regarded as a whole-tree operation, ETE allows to root subparts of the tree without affecting to its parent tree structure.

from ete2 import Tree
t = Tree('(((A,C),((H,F),(L,M))),((B,(J,K))(E,D)));')
print "Original tree:"
print t
#                              /-A
#                    /--------|
#                   |          \-C
#                   |
#          /--------|                    /-H
#         |         |          /--------|
#         |         |         |          \-F
#         |          \--------|
#         |                   |          /-L
#         |                    \--------|
#---------|                              \-M
#         |
#         |                    /-B
#         |          /--------|
#         |         |         |          /-J
#         |         |          \--------|
#          \--------|                    \-K
#                   |
#                   |          /-E
#                    \--------|
#                              \-D
#
# Each main branch of the tree is independently rooted.
node1 = t.get_common_ancestor("A","H")
node2 = t.get_common_ancestor("B","D")
node1.set_outgroup("H")
node2.set_outgroup("E")
print "Tree after rooting each node independently:"
print t
#
#                              /-F
#                             |
#                    /--------|                    /-L
#                   |         |          /--------|
#                   |         |         |          \-M
#                   |          \--------|
#          /--------|                   |          /-A
#         |         |                    \--------|
#         |         |                              \-C
#         |         |
#         |          \-H
#---------|
#         |                    /-D
#         |          /--------|
#         |         |         |          /-B
#         |         |          \--------|
#          \--------|                   |          /-J
#                   |                    \--------|
#                   |                              \-K
#                   |
#                    \-E

Working with branch distances

The branch length between one node an its parent is encoded as the TreeNode.dist attribute. Together with tree topology, branch lengths define the relationships among nodes.

Getting distances between nodes

The TreeNode.get_distance() method can be used to calculate the distance between two connected nodes. There are two ways of using this method: a) by querying the distance between two descendant nodes (two nodes are passed as arguments) b) by querying the distance between the current node and any other relative node (parental or descendant).

from ete2 import Tree

# Loads a tree with branch lenght information. Note that if no
# distance info is provided in the newick, it will be initialized with
# the default dist value = 1.0
nw = """(((A:0.1, B:0.01):0.001, C:0.0001):1.0,
(((((D:0.00001:0,I:0):0,F:0):0,G:0):0,H:0):0,
E:0.000001):0.0000001):2.0;"""
t = Tree(nw)
print t
#                              /-A
#                    /--------|
#          /--------|          \-B
#         |         |
#         |          \-C
#         |
#         |                                                  /-D
#         |                                        /--------|
#---------|                              /--------|          \-I
#         |                             |         |
#         |                    /--------|          \-F
#         |                   |         |
#         |          /--------|          \-G
#         |         |         |
#          \--------|          \-H
#                   |
#                    \-E
#
# Locate some nodes
A = t&"A"
C = t&"C"
# Calculate distance from current node
print "The distance between A and C is",  A.get_distance("C")
# Calculate distance between two descendants of current node
print "The distance between A and C is",  t.get_distance("A","C")
# Calculate the toplogical distance (number of nodes in between)
print "The number of nodes between A and D is ",  \
    t.get_distance("A","D", topology_only=True)

Additionally to this, ETE incorporates two more methods to calculate the most distant node from a given point in a tree. You can use the TreeNode.get_farthest_node() method to retrieve the most distant point from a node within the whole tree structure. Alternatively, TreeNode.get_farthest_leaf() will return the most distant descendant (always a leaf). If more than one node matches the farthest distance, the first occurrence is returned.

Distance between nodes can also be computed as the number of nodes between them (considering all branch lengths equal to 1.0). To do so, the topology_only argument must be set to True for all the above mentioned methods.

# Calculate the farthest node from E within the whole structure
farthest, dist = (t&"E").get_farthest_node()
print "The farthest node from E is", farthest.name, "with dist=", dist
# Calculate the farthest node from E within the whole structure,
# regarding the number of nodes in between as distance value
# Note that the result is differnt.
farthest, dist = (t&"E").get_farthest_node(topology_only=True)
print "The farthest (topologically) node from E is", \
    farthest.name, "with", dist, "nodes in between"
# Calculate farthest node from an internal node
farthest, dist = t.get_farthest_node()
print "The farthest node from root is is", farthest.name, "with dist=", dist
#
# The program results in the following information:
#
# The distance between A and C is 0.1011
# The distance between A and C is 0.1011
# The number of nodes between A and D is  8.0
# The farthest node from E is A with dist= 1.1010011
# The farthest (topologically) node from E is I with 5.0 nodes in between
# The farthest node from root is is A with dist= 1.101

getting midpoint outgroup

In order to obtain a balanced rooting of the tree, you can set as the tree outgroup that partition which splits the tree in two equally distant clusters (using branch lengths). This is called the midpoint outgroup.

The TreeNode.get_midpoint_outgroup() method will return the outgroup partition that splits current node into two balanced branches in terms of node distances.

from ete2 import Tree
# generates a random tree
t = Tree();
t.populate(15);
print t
#
#
#                    /-qogjl
#          /--------|
#         |          \-vxbgp
#         |
#         |          /-xyewk
#---------|         |
#         |         |                    /-opben
#         |         |                   |
#         |         |          /--------|                    /-xoryn
#          \--------|         |         |          /--------|
#                   |         |         |         |         |          /-wdima
#                   |         |          \--------|          \--------|
#                   |         |                   |                    \-qxovz
#                   |         |                   |
#                   |         |                    \-isngq
#                    \--------|
#                             |                    /-neqsc
#                             |                   |
#                             |                   |                              /-waxkv
#                             |          /--------|                    /--------|
#                             |         |         |          /--------|          \-djeoh
#                             |         |         |         |         |
#                             |         |          \--------|          \-exmsn
#                              \--------|                   |
#                                       |                   |          /-udspq
#                                       |                    \--------|
#                                       |                              \-buxpw
#                                       |
#                                        \-rkzwd
# Calculate the midpoint node
R = t.get_midpoint_outgroup()
# and set it as tree outgroup
t.set_outgroup(R)
print t
#                              /-opben
#                             |
#                    /--------|                    /-xoryn
#                   |         |          /--------|
#                   |         |         |         |          /-wdima
#                   |          \--------|          \--------|
#          /--------|                   |                    \-qxovz
#         |         |                   |
#         |         |                    \-isngq
#         |         |
#         |         |          /-xyewk
#         |          \--------|
#         |                   |          /-qogjl
#         |                    \--------|
#---------|                              \-vxbgp
#         |
#         |                    /-neqsc
#         |                   |
#         |                   |                              /-waxkv
#         |          /--------|                    /--------|
#         |         |         |          /--------|          \-djeoh
#         |         |         |         |         |
#         |         |          \--------|          \-exmsn
#          \--------|                   |
#                   |                   |          /-udspq
#                   |                    \--------|
#                   |                              \-buxpw
#                   |
#                    \-rkzwd