dendropy.utility.error.UltrametricityError when using TACT

[Cactusolo]

Hi @jonchang ,

I'm trying to use TACT adding some species to GBMB.tre from Smith and Brown (2018). The backbone is fine, which was confirmed by phyx.

[cactus]$ pxlstr -t GBMB_bi-rmbadtip.tre [some tips are cleaned]
tree #: 0
rooted: true
binary: true
nterminal: 78929
ninternal: 78928
branch lengths: true
rttipvar: 2.23112e-10
treelength: 490017
ultrametric: true
rootheight: 325.05

When I run TACT with tact_add_taxa function, it gives error as below:

Traceback (most recent call last):
File "/home/cactus/.local/bin/tact_add_taxa", line 8, in
sys.exit(main())
File "/home/cactus/.local/pipx/venvs/tact/lib/python3.7/site-packages/click/core.py", line 829, in call
return self.main(*args, **kwargs)
File "/home/cactus/.local/pipx/venvs/tact/lib/python3.7/site-packages/click/core.py", line 782, in main
rv = self.invoke(ctx)
File "/home/cactus/.local/pipx/venvs/tact/lib/python3.7/site-packages/click/core.py", line 1066, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "/home/cactus/.local/pipx/venvs/tact/lib/python3.7/site-packages/click/core.py", line 610, in invoke
return callback(*args, **kwargs)
File "/home/cactus/.local/pipx/venvs/tact/lib/python3.7/site-packages/tact/cli_add_taxa.py", line 392, in main
tree.calc_node_ages()
File "/home/cactus/.local/pipx/venvs/tact/lib/python3.7/site-packages/dendropy/datamodel/treemodel.py", line 5666, in calc_node_ages
subtree=subtree,
dendropy.utility.error.UltrametricityError: Tree is not ultrametric within threshold of 1e-05: 1.0000000017384991e-05.
Encountered in subtree of node <Node object at 0x7fc312709470: 'None' (None)> (edge length of 1.291905):

####huge newick tree skipped ...####

Age of children:

• <Node object at 0x7fc3127094e0: 'None' (None)>: has age of 88.17562399999998 and edge length of 5.093615, resulting in parent node age of 93.26923899999998
• <Node object at 0x7fc3126849b0: 'None' (None)>: has age of 25.721769000000002 and edge length of 67.54748, resulting in parent node age of 93.269249

I suspect it could be issues from dendropy, not really TACT, but would like to hear from your advice first. Thank you!

Happy to mail you my tree file and taxonomy data if you have time to check it.

Cheers,

Miao

[jonchang]

ape, DendroPy, and phyx all compute ultrametricity in different ways. I disagree with DendroPy's approach since it looks at every single node in the tree and raises an error if the ages of the left and right children of the parent node differ by some threshold (set to 1e-5). I think ape looks at the relative maximum difference in root-to-tip distance and phyx looks at the variance in root-to-tip distances (which ape used to use).

I'll probably write something in tact to address this but for now you can use nnls.tree(cophenetic(phy), phy, TRUE) in the R package phangorn to fix the tree.

[Cactusolo]

Thanks for your respond @jonchang ! I have tried with phytools::force.ultramtric() with "method = 'extend'" and it requires 45 Gb memory request; with "method = 'nnls'" and phangorn::nnls.tree(cophenetic(phy), phy, TRUE) raising the same error msg:

> library("phangorn")
Loading required package: ape
> tree <- read.tree("GBMB_bi-rmbadtip.tre")
> library("phytools")
Loading required package: maps
> tree2 <- force.ultrametric(tree, method="nnls")
Error in double(nm * nm) : vector size cannot be NA
In addition: Warning message:
In nm * nm : NAs produced by integer overflow
> tree3 <- nnls.tree(cophenetic(tree), tree, rooted =TRUE)
Error in double(nm * nm) : vector size cannot be NA
In addition: Warning message:
In nm * nm : NAs produced by integer overflow

But no worries, I'm looking forward to your update.

Thank you so much!
Miao

[jonchang]

Miao, can you see if #231 fixes the ultrametricity check?

[Cactusolo]

Hi, I'm trying now. Should I re-install tact and rerun again?

I'm on a Ubuntu machine.

I have upgraded TACT:

[cactus]$ pipx upgrade tact
upgraded package tact from 0.3.1 to 0.3.4 (location: /home/cactus/.local/pipx/venvs/tact)

And reran it, and the issue remains.

Miao

[jonchang]

You'll need to install it from the branch, like so:

pipx install git+https://github.com/jonchang/tact.git@fix-ultrametricity

[Cactusolo]

Hi,

Thanks for the tip! I have upgraded the tact. Now the error message is new, said:

[cactus]$ cat tacted.test.log.txt
Reading taxonomy
Reading backbone
Tree is not ultrametric!

However, phyx said the tree is ultrametric:

[cactus]$ pxlstr -t GBMB_bi-rmbadtip.tre
tree #: 0
rooted: true
binary: true
nterminal: 78929
ninternal: 78928
branch lengths: true
rttipvar: 2.23112e-10
treelength: 490017
ultrametric: true
rootheight: 325.05

Thanks!

Miao

[jonchang]

Can you email me the tree?

[Cactusolo]

Can you email me the tree?

Yes, I send the data to you at "me@jonathanchang.org" [not sure about the email address :P].

Thanks!

Miao

[jonchang]

Tree is not ultrametric: (('Platea latifolia', 325.0477049999999), ('Ginkgo biloba', 325.05001799999997))

I think I'll solve this by just letting you set an ultrametricity precision.

[jonchang]

Ok, try pipx uninstall tact && pipx install git+https://github.com/jonchang/tact.git@fix-ultrametricity again and try passing --ultrametricity-precision. Note that this will coerce your ages to be a bit older (so the root-to-tip distance for Platea latifolia will be slightly larger).

[Cactusolo]

Hey @jonchang , I have done as what you instructed pipx uninstall tact && pipx install git+https://github.com/jonchang/tact.git@fix-ultrametricity, and pass --ultrametricity-precision.

please let me know if I did anything wrong here, the program has not finished with error msg:

Tree is not ultrametric!
Platea latifolia has a root distance of 325.0477049999999, but Ginkgo biloba has 325.05001799999997
If this is unexpected, consider setting --ultrametricity-precision or using phytools::force.ultrametric in R
allTaxoPlantsTrees.gbmb.tacted.test.log.txt (END)

tact_add_taxa --backbone backbone.tre --taxonomy taxonomy.ed.tre --output tacted.test --ultrametricity-precision 1e-6

[jonchang]

Try with a higher number. 1e-6 implies six figs of similarity, so you'd need something like 1e-3 or 1e-2.

[Cactusolo]

Try with a higher number. 1e-6 implies six figs of similarity, so you'd need something like 1e-3 or 1e-2.

Thanks for the detailed instruction @jonchang !! it worked and finished this time. 👍

[jonchang]

This fix is now released as part of tact 0.4.0: https://github.com/jonchang/tact/releases/tag/v0.4.0