partitioning in BEAUti 1.5.4

[rojasmartin.cu]

Dear colleagues:

I am dealing with 55 sequences (at the genus level) of 5 genes
(nuclear and mitochondrial). I am not able to partitioning the
sequences with BEAUti 1.5.4 in order to obtain a single tree for all
markers (and not a tree for each), allowing each marker to follow its
own substitution model. The xml file I get is very different from that
made by hand after generating the xml file in BEAUti 1.4.8 and then
partitioning with the suggestions found in
http://sites.google.com/site/tlpcouvreur/beastpartitioning. I am not
able to follow the tutorial Divergence Dating (Primates) v1.1a. Could
anyone help me?

Thanks in advance.

Danny.

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To post to this group, send email to beast...@googlegroups.com.
To unsubscribe from this group, send email to beast-users...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/beast-users?hl=en.

[julien]

Hi Danny,

This completely normal... The file by Denis Couvreur assumes a single
clock (1 UCLD & 1 meanrate), whose rate is split among genes (mu
parameters) according with their respective length.

The file built by Beauty 1.5.4 can assume as many clock as genes: so
you have several UCLD and meanrates.

You can implement the model you prefer ! The one set in the file of
Denis Couvreur can also be built from Beauty 1.5.4: you need to relax
substitution models without relaxing clocks nor trees.

Hope that helps !

Julien

On 20 avr, 18:41, "rojasmartin.cu" <rojasmartin...@gmail.com> wrote:
> Dear colleagues:
>
> I am dealing with 55 sequences (at the genus level) of 5 genes
> (nuclear and mitochondrial). I am not able to partitioning the
> sequences with BEAUti 1.5.4 in order to obtain a single tree for all
> markers (and not a tree for each), allowing each marker to follow its
> own substitution model. The xml file I get is very different from that
> made by hand after generating the xml file in BEAUti 1.4.8 and then

> partitioning with the suggestions found inhttp://sites.google.com/site/tlpcouvreur/beastpartitioning. I am not

[Tod Reeder]

Julien has described the unlinking of the clocks, but it sounds like
Danny's issue involves how to set up the partitioned analysis without
having individual trees being inferred for each partition (i.e.,
"unlinked" trees). For the partitioned analysis, in BEAUTi, you need
to unlink the substitution models and the clock models among
partitions. However, for the trees (topologies), if I understand the
method of concatenation in BEAUTi, then you need to "link" the trees;
that way you get a single tree estimated for all the data which will
be analyzed with models specific to each partition since they are
"unlinked". Hope this helps (and hope it is correct).

[rojasmartin.cu]

Hi Julien:

Thank you very much for your help. Now things are clearer to me. But I
still have a couple of questions.

When keeping unlinked the clock and the trees in BEAUti 1.5.4 I have
observed that it is assumed the first partition. Then unlinking allows
to select the clock and the tree of other partition and then linking
again makes this new clock and tree as the single one for the matrix.
Should I left by default the clock and the tree of the first partition
or if I can change it what criteria should I follow (perhaps to choose
the most informative or the larger partition)?

The second question: which of both procedures (partitioning with one
single clock and tree for all the genes or allowing each gene its own
clock and tree) allows to obtain the most accurate and precise
divergence time estimations?

Thanks again for your help.

Merci beaucoup.

Danny.

[rojasmartin.cu]

Thank you Tod for your help.

Indeed I am intending to have a single tree for all data instead of
individual trees for each partition. I keep the two questions I posted
before to Julien: (1) which partition to use when linking trees and
(2) what is the best thing (in terms of accuracy and precision) to do:
unlink trees or keep the trees (and the clocks?) linked?

[julien]

You can look at your xml file and see that models components are
always applied to right "patterns id"..."" if unlinked or to the whole
"alignment id="..."" if linked... So I think that the partition name
that appears, while the substitution model and/or clocks are linked,
does not matter actually.

Obviously you shoudn't relax trees among genes, unless having several
specimens per species so to use variance in coalescence events (among
genes) to accurate your molecular dating (the so called "Star-Beast"
method )...

Usually I relax substitution models and clocks, except if data are too
poorly informative (use only one clock in this case)...

Hope that helps !