Calculating bootstrap support values for an existing tree

[Nick Croucher]

Hi RAxML team,

Thanks very much for all your work on RAxML and RAxML-NG over the years!

A difference in the behaviour of RAxML and RAxML-NG has been brought to my attention, and I was wondering whether you might be able to explain the cause?

I have a starting tree: ST239.final_tree.tre

I have a set of bootstrap trees: RAxML_bootstrap.ST239.iteration_3.bootstrapped_trees

If I annotate support on the former using the latter using RAxML v8.2.12, I get the expected tree returned (Newick here):
raxmlHPC-SSE3 -safe -m GTRCAT --JC69 -p 5730 -f b -t ST239.final_tree.tre -z RAxML_bootstrap.ST239.iteration_3.bootstrapped_trees -n raxml_bootstrap

If use the same inputs with RAxML-NG 1.0.1 or 1.0.3, then the branch on which the original tree was rooted is longer than expected (Newick here) - one side of it appears to have ~doubled in length:
raxml-ng --support --bs-trees RAxML_bootstrap.ST239.iteration_3.bootstrapped_trees --tree ST239.final_tree.tre --prefix raxmlng_bootstrap

I appreciate rooting is not of much relevance with bootstraps, but I wasn't expecting this alteration in behaviour. Any help understanding these differences would be much appreciated!

[Grimm]

Hi Nick,

this is just a optical phenomenon because the two trees are differently rooted when opened in a tree viewer.

In fact, they are trees are identical. This is the tanglegram of the two, when midpoint-rooted (rooted at the longest branch).

Note that rooting is a purely graphical thing, and can mislead our perception when we look at two differently rooted versions of the same tree (why I personally always used the actually inferred, unrooted tree)

Cheers, Guido

[Alexandros Stamatakis]

Thanks Guido,

Here is a paper on discussing the issue Guido describes.

https://academic.oup.com/mbe/article/34/6/1535/3077051?login=true

Alexis

On 21.10.21 14:17, Grimm wrote:
> Hi Nick,
>
> this is just a optical phenomenon because the two trees are differently
> rooted when opened in a tree viewer.
>
> In fact, they are trees are identical. This is the tanglegram of the
> two, when midpoint-rooted (rooted at the longest branch).
>
> Note that rooting is a purely graphical thing, and can mislead our
> perception when we look at two differently rooted versions of the same
> tree (why I personally always used the actually inferred, unrooted tree)
>
> Cheers, Guido
>

> Midpoint.jpg
>
> Nick Croucher schrieb am Donnerstag, 21. Oktober 2021 um 13:49:30 UTC+2:
>
> Hi RAxML team,
>
> Thanks very much for all your work on RAxML and RAxML-NG over the years!
>
> A difference in the behaviour of RAxML and RAxML-NG has been brought
> to my attention, and I was wondering whether you might be able to
> explain the cause?
>
> I have a starting tree: ST239.final_tree.tre

> <https://drive.google.com/file/d/1SZGSiN3iMigL7TTBUTiSDF3dEbSLHUCf/view?usp=sharing>

>
> I have a set of bootstrap trees:
> RAxML_bootstrap.ST239.iteration_3.bootstrapped_trees

> <https://drive.google.com/file/d/1_bUEJ6QpJ-57pWi_7fcSevEAGaC83kg2/view?usp=sharing>

>
> If I annotate support on the former using the latter using RAxML
> v8.2.12, I get the expected tree returned (Newick here

> <https://drive.google.com/file/d/1SyiSXNVU9wkJr2X0fZ276SkGIK_3m5O7/view?usp=sharing>):

> raxmlHPC-SSE3 -safe -m GTRCAT --JC69 -p 5730 -f b -t
> ST239.final_tree.tre -z
> RAxML_bootstrap.ST239.iteration_3.bootstrapped_trees -n raxml_bootstrap
>
> If use the same inputs with RAxML-NG 1.0.1 or 1.0.3, then the branch
> on which the original tree was rooted is longer than expected
> (Newick here

> <https://drive.google.com/file/d/1X6OyR7Rko5egJjhBr84HJqu37k1XORqU/view?usp=sharing>)

> - one side of it appears to have ~doubled in length:
> raxml-ng --support --bs-trees
> RAxML_bootstrap.ST239.iteration_3.bootstrapped_trees --tree
> ST239.final_tree.tre --prefix raxmlng_bootstrap
>
> I appreciate rooting is not of much relevance with bootstraps, but I
> wasn't expecting this alteration in behaviour. Any help
> understanding these differences would be much appreciated!
>

> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/raxml/19165713-bffa-4bc7-a61c-a2e7ac7774ffn%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/19165713-bffa-4bc7-a61c-a2e7ac7774ffn%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
Affiliated Scientist, Evolutionary Genetics and Paleogenomics (EGP) lab,
Institute of Molecular Biology and Biotechnology, Foundation for
Research and Technology Hellas

www.exelixis-lab.org

[Nick Croucher]

That's really helpful, thank you. I could see the topologies were the same when midpoint rooted, but failed to get my head around the shift in the root position within the longest branch between the two outputs. Thanks very much to you both for your patience and help!