Interpreting RAxML output for Internode/Tree Certainty calculations

[Panos Ioannidis]

Hi!

I want to build a tree for my organism of interest, based on concatenation of several genes. So in order to select what genes give me the most information I decided to first build a tree for each gene, individually. The next step is to calculate the certainty of each of these calculating (a) the average bootstrap support and (b) the "internode/tree certainty" values (as per Salichos and Rokas, 2013).

I started playing with (b) and have a question, regarding interpretation of the output.

So I'm using the bootstrap trees of each gene tree to calculate the tree certainty, using the following the command

raxmlHPC-PTHREADS-AVX -L MR -z <bootstrap trees> -m GTRCAT -n T1 -T4

The part of the output I don't really understand is the following

Tree certainty for this tree: 5.282070

Relative tree certainty for this tree: 0.660259

Tree certainty including all conflicting bipartitions (TC-All) for this tree: 5.282070

Relative tree certainty including all conflicting bipartitions (TC-All) for this tree: 0.660259

I'm suspecting that I should use the "relative tree certainty", but I'm not sure which one; what's the difference between TC and TC-All? Also, is the relative tree certainty going from 0 (no certainty at all) to 1 (no conflicts)? Are there cases were I'm supposed to use the absolute (?) tree certainty values?

Thanks,

Panos

[Alexandros Stamatakis]

Dear Pano,

Firstly, please find attached a manual describing the RAxML
implementation of the TC/IC calculations.

> I want to build a tree for my organism of interest, based on concatenation
> of several genes. So in order to select what genes give me the most
> information I decided to first build a tree for each gene, individually.
> The next step is to calculate the certainty of each of these calculating
> (a) the average bootstrap support and (b) the "internode/tree certainty"
> values (as per Salichos and Rokas, 2013).
>
> I started playing with (b) and have a question, regarding interpretation of
> the output.
>
> So I'm using the bootstrap trees of each gene tree to calculate the tree
> certainty, using the following the command
>
> raxmlHPC-PTHREADS-AVX -L MR -z <bootstrap trees> -m GTRCAT -n T1 -T4

Looks good, you may also consider using the extended majority rule tree
(-L MRE).

> The part of the output I don't really understand is the following
>
> Tree certainty for this tree: 5.282070
> Relative tree certainty for this tree: 0.660259
>
> Tree certainty including all conflicting bipartitions (TC-All) for this
> tree: 5.282070
> Relative tree certainty including all conflicting bipartitions (TC-All) for
> this tree: 0.660259
>
> I'm suspecting that I should use the "relative tree certainty", but I'm not
> sure which one;

This is just something additional I implemented, that was not described
by Antonis and Leonidas.

> what's the difference between TC and TC-All?

TC is the sum of all IC scores
TC-All is the sum of all IC-All scores.

I believe it's described in the paper, but basically the IC-All scores
is computed by taking all conflicting bipartitions into account that
have >= 5% support and not only the most supported conflicting bipartition.

> Also, is the
> relative tree certainty going from 0 (no certainty at all) to 1 (no
> conflicts)?

Yes, that's correct.

> Are there cases were I'm supposed to use the absolute (?) tree
> certainty values?

I can't think of any, I believe though, that the relative TC scores are
more meaningful because they have been normalized.

Cheers,

Alexis

>
> Thanks,
> Panos
>

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
Adjunct Professor, Dept. of Ecology and Evolutionary Biology, University
of Arizona at Tucson

www.exelixis-lab.org

[Panos Ioannidis]

Hi Alexi,

That was really helpful.

Thanks a lot!

Panos