Majority rule consensus tree

[Christian A]

Hi,

I searched this forum for this issue, but I didn't really find a
satisfying answer...

I am confused about how to build majority rule consensus tree with
RAxML... When building a majority rule consensus tree after my ML
analysis, I have support values lower than 50 in my tree (down to 0).
How is this possible if this is supposed to be a majority rule
consensus tree? What I want (and what other programs, like MrBayes,
for example) do is to introduce a polytomy if support values are less
than 50%, based on a set of trees.

I used the following command in RAxML:
raxmlHPC -m GTRGAMMA -J MR -f b -t RAxML_result.optimize -z
RAxML_bootstrap.bootstrap -n consensus

So, what I want is to take the ML tree (RAxML_result.optimize), and
use the boostrap trees to make support values for each internal node
in the ML tree. If a particular node is supported by less than 50%,
the ML tree topology is altered and a polytomy is introduced. Thus,
at the end, only the strongly supported nodes are still present,
whereas all weakly supported nodes were collapsed.

Thanks,
Christian

[Akifumi S. Tanabe]

Hi Christian,

To make majority rule consensus tree, you don't need specify ML tree
file via "-t". You should just type the following command.
raxmlHPC -f b -m GTRGAMMA -J MR -z RAxML_bootstrap.bootstrap -n consensus

If you give ML tree file via "-t", RAxML counts the times of emergence
of the bipartitions which is contained in the ML tree, and mapped them as
support values to the ML tree.

On Fri, 6 Jan 2012 06:28:12 -0800 (PST)
Christian A <chrar...@googlemail.com> wrote in
<4aa872f6-8bd2-431a...@j10g2000vbe.googlegroups.com>

> So, what I want is to take the ML tree (RAxML_result.optimize), and
> use the boostrap trees to make support values for each internal node
> in the ML tree. If a particular node is supported by less than 50%,
> the ML tree topology is altered and a polytomy is introduced. Thus,
> at the end, only the strongly supported nodes are still present,
> whereas all weakly supported nodes were collapsed.

It's not majority rule consensus tree. RAxML cannot compute it, and much
the same is true on MrBayes as far as I know.

--
Akifumi S. Tanabe <akifumi...@gmail.com>

[Alexis]

Thanks Akifumi,

There are two different things here:

1. Draw the BS support from the bootstrap trees on the ML tree using
the -f b option. This will just count
how frequently bipartitions of the ML tree occur in the bootstrap
replicate trees and requires you to pass
a set of BS trees as well as the ML tree (or any reference tree for
that matter) to the program.

2. Build some sort of consensus tree from a collection of trees with -
J MR or -J MRE or -J STRICT this requires
you to just pass one file containing several trees to RAxML and if you
use -J MR will indeed just show branches with more than 50% support.

Hope this helps,

Alexis

On 6 Jan., 17:03, "Akifumi S. Tanabe" <akifumi.tan...@gmail.com>
wrote:

> Hi Christian,
>
>   To make majority rule consensus tree, you don't need specify ML tree
> file via "-t". You should just type the following command.
> raxmlHPC -f b -m GTRGAMMA -J MR -z RAxML_bootstrap.bootstrap -n consensus
>
>   If you give ML tree file via "-t", RAxML counts the times of emergence
> of the bipartitions which is contained in the ML tree, and mapped them as
> support values to the ML tree.
>
> On Fri, 6 Jan 2012 06:28:12 -0800 (PST)

> Christian A <chrarnol...@googlemail.com> wrote in
> <4aa872f6-8bd2-431a-a8f5-40e27ff8a...@j10g2000vbe.googlegroups.com>

>
> > So, what I want is to take the ML tree (RAxML_result.optimize), and
> > use the boostrap trees to make support values for each internal node
> > in the ML tree. If a particular node is supported by less than 50%,
> > the ML tree topology is altered and a polytomy is introduced.  Thus,
> > at the end, only the strongly supported nodes are still present,
> > whereas all weakly supported nodes were collapsed.
>
>   It's not majority rule consensus tree. RAxML cannot compute it, and much
> the same is true on MrBayes as far as I know.
>
> --

> Akifumi S. Tanabe <akifumi.tan...@gmail.com>

[Jeff]

A quick question related to this thread regarding consensus trees.  Is this method appropriate for multilocus data (i.e., many trees based on many different genes)?  If yes, great!  If no, can you recommend a means of accomplishing this task?  Thanks!

[Andre J. Aberer]

Hi Jeff,

> A quick question related to this thread regarding consensus trees. Is
> this method appropriate for multilocus data (i.e., many trees based on
> many different genes)? If yes, great! If no, can you recommend a
> means of accomplishing this task? Thanks!

I have limited experience with that, but what I know, consensus trees
based on such trees often show poor resolution (but why not give it a
try?).

In this context, as an alternative sometimes the "Maximum Agreement
Subtree" is used. This is the maximal (in terms of taxa) fully
bifurcating subtree that is contained in every tree in your
set. Usually, there exist many of them, which makes interpretation
challenging. You can find an implementation of the MAST in PAUP or
RogueNaRok (https://github.com/aberer/RogueNaRok).

--
Best regards,
Andre J. Aberer

M.Sc. (Bioinformatics)
Scientific Computing Group

Heidelberg Institute for Theoretical Studies (HITS gGmbH)
Schloss-Wolfsbrunnenweg 35
D-69118 Heidelberg

Tel.: +49 6221 533 264
Fax: +49 6221 533 298
Email: andre....@h-its.org
WWW: http://www.exelixis-lab.org
http://www.h-its.org/english/research/sco/index.php

Amtgericht Mannheim / HRB 337446
Managing Directors: Dr. h.c. Dr.-Ing. E.h. Klaus Tschira, Prof. Dr.-Ing. Andreas Reuter

[Jeff]

Thanks Andre,

I've tried the consensus tree and it appears to be in agreement with the MEGA concatenated tree.  I'm hoping for a method of accounting for incongruence among genes that yields an alternate phylogeny.  I like your suggestion to use MAST in PAUP but I've yet to put up the $150 for the software and I'm reluctant to do so; however, I also want to try Zonal Phylogeny which required PAUP so I may buy it.  I'm also exploring using PhyloNet for this purpose but having some initial difficulty.  When I find a solution I'll follow up on this thread.

[Andre J. Aberer]

Hi Jeff,

> I've tried the consensus tree and it appears to be in agreement with
> the MEGA concatenated tree. I'm hoping for a method of accounting for
> incongruence among genes that yields an alternate phylogeny.

Okay, so for this specific task the MAST may not be the best tool. How
about a clustering of RF-distances of the trees. Maybe there are
well-defined clusters of gene trees, that then can be summarized as
consensus trees.

> I like your suggestion to use MAST in PAUP but I've yet to put up the
> $150 for the software and I'm reluctant to do so; however, I also want
> to try Zonal Phylogeny which required PAUP so I may buy it. I'm also
> exploring using PhyloNet for this purpose but having some initial
> difficulty. When I find a solution I'll follow up on this thread.

As said, we also provide a free implementation of the MAST in RogueNaRok
(called rnr-mast).

[Jeff]

Hi Andre,

I concur that the consensus method is crude.  Thanks for the link to RogueNaRok - I completely missed that in the previous post.  I'll try it out!  I'm also exploring BUCKy for this purpose.

[Jed]

> I've tried the consensus tree and it appears to be in agreement with
> the MEGA concatenated tree.  I'm hoping for a method of accounting for
> incongruence among genes that yields an alternate phylogeny.  

You could try SplitsTree, which provides an output as a network, rather than a bifurcating or unresolved tree. You can use it to visualize alternatives, and even (I think) evaluate support.

Jed