Conversion from ete3 trees to skbio trees

[mortonjt]

Right now, there is no way (I know of) that can convert ete3 trees into skbio trees. So right now, I'm doing the conversions by passing around the newick strings. However, this is showing to be problematic. Specifically if I try to run the following code

import random
from ete3 import Tree
from skbio import TreeNode

random.seed(0)
t = Tree()
t.populate(50, random_branches=True)
for i, n in enumerate(t.traverse()):
    n.name = 'y%d' % i
st = TreeNode.read([t.write()])

I get the following two trees

# ETE3
>>> t.children
[Tree node 'y1' (-0x7fffffffedd8a115), Tree node 'y2' (-0x7fffffffedd8a107)]
# skbio
>>> st.children
[<TreeNode, name: 0.139193, internal node count: 33, tips count: 35>,
 <TreeNode, name: 0.460504, internal node count: 13, tips count: 15>]

So the ete3 trees have a different naming scheme that I don't quite understand.
The hack around this for the time being is just to fix the resulting skbio tree.

st = TreeNode.read([t.write()])
for i, n in enumerate(st.levelorder(include_self=True)):
    n.name = 'y%d' % i 

However, it would be nice if there were a formal solution available. cc @gregcaporaso

[jhcepas]

@mortonjt, you need to export the newick string using format=1, so it keeps internal node names.

import random
from ete3 import Tree
from skbio import TreeNode
from io import StringIO

random.seed(0)
t = Tree()
t.populate(10, random_branches=True)
for i, n in enumerate(t.traverse()):
    n.name = 'y%d' % i
print(t.get_ascii(attributes=["name", "dist"]))
print(t.children)

st = TreeNode.read(StringIO(t.write(format=1, format_root_node=True)))
print(st.ascii_art())
print(st.children)

#                                                    /-y7, 0.09327186435241008
#                              /y3, 0.5046868558173903
#                             |                      \-y8, 0.8400912165554131
#        /y1, 0.7579544029403025
#       |                     |                                             /-y15, 0.6118970848141451
#       |                     |                       /y9, 0.9872592010330129
#       |                      \y4, 0.28183784439970383                     \-y16, 0.8280632784038988
#       |                                            |
# -y0, 1.0                                            \-y10, 0.5325636340885271
#       |
#       |                                           /-y11, 0.3101475693193326
#       |                     /y5, 0.9182343317851318
#       |                    |                     |                       /-y17, 0.5598136790149003
#       |                    |                      \y12, 0.7298317482601286
#        \y2, 0.420571580830845                                            \-y18, 0.35379132951924075
#                            |
#                            |                      /-y13, 0.9666063677707588
#                             \y6, 0.8298529036589914
#                                                   \-y14, 0.47700977655271704
# [Tree node 'y1' (0x100612a9), Tree node 'y2' (-0x7fffffffeff9ed54)]
#                               /-y7
#                     /y3------|
#                    |          \-y8
#           /y1------|
#          |         |                    /-y15
#          |         |          /y9------|
#          |          \y4------|          \-y16
#          |                   |
# -y0------|                    \-y10
#          |
#          |                    /-y11
#          |          /y5------|
#          |         |         |          /-y17
#          |         |          \y12-----|
#           \y2------|                    \-y18
#                    |
#                    |          /-y13
#                     \y6------|
#                               \-y14
# [<TreeNode, name: y1, internal node count: 3, tips count: 5>, <TreeNode, name: y2, internal node count: 3, tips count: 5>]

[mortonjt]

Awesome! I think that basically resolves my immediate issues. Thanks!

However, it looks like the attributes aren't copied over from ete3 to skbio.
Copying these attributes over could be still be useful in a conversion function.

[jhcepas]

nope, attributes are not transferred through serialization. I discussed this with @gregcaporaso et al once, but skbio.Tree was still not stable and we were unsure about how to transfer attrs.
If done, it would probably make sense to have method that copies the tree properly and not through newick serialization.

[jhcepas]

closing as exporting features does not seem to be necessary for the moment.