species tree inference with no calibration

[elena]

Hi all,
We would like to infer the species tree based on 3 phased nuclear and 2 mitochondrial loci. We are having problems with convergence and I would like to use starbeast to infer the species tree but without necessarily calibrate the tree. Do you have any clue about which option I should use in choosing the clock model priors.
Thanks

[Diogo Silva]

Hello Elena

If you're not interested in calibrating your tree in geological time
units, then you should fix the molecular rate of your clock model to
"1". In this way, your tree will be in subst/site (like a MrBayes
tree) rather than subst/site/time.

Hope it helps
Cheers
Diogo

[elena]

Hi Diogo,

Thank you!
I am not sure whether, in case of no convergence, it would be a good practice to use starbeast only to infer the species tree and use beast to calibrate the nodes based on a particular gene or set of genes with known substitution rates (i.e. mitochondrial).
What do you think?

cheers
elena

[Diogo Silva]

The use of BEAST or *BEAST will depend on your dataset. However, if
your dataset is appropriate for *BEAST (i.e., multiple species and
multiple individuals per species, multiple loci), based on my
experience and on the literature, I would recommend you to use it to
infer the species tree and the TMRCA of your nodes of interest. *BEAST
uses the coalescent theory to take into account several issues related
to the stochasticity of the process of lineage sorting, demography and
divergence times (most of these are very well demonstrated on
McCormack et al., 2010, Evolution), which BEAST does not (Also read
Heled and Drummond, 2010, MBE). Moreover, if you have detected some
discordance between your individual gene trees, that's even more
reason to use *BEAST since it takes into consideration the probability
of unsorted ancestral polimorphisms. Overall, *BEAST will tend to give
you more recent estimates of divergence times and, more importantly,
speciation times.

Regarding your lack of convergence, it is very difficult to provide an
opinion without knowing the cause. Could you specify which parameters
are not converging and provide a more detailed picture of your input
XML file?

Cheers
Diogo

[Suzanne Williams]

Hi Diogo,

I was very interested to see that you have identified that species
divergences seem to be younger in *BEAST analyses than BEAST analyses. We
have found a similar pattern with our data.

Do you know whether this finding has been published somewhere (so we can
cite it)? And do you have any idea why this is occurring?

Thanks for your help.

Suzanne Williams

--------------------------------------------------------
Dr Suzanne Williams
Zoology Dept
Natural History Museum
Cromwell Rd
London SW7 5BD
United Kingdom
Tel: + 44 (0) 207 942 5351 (office) 5774 (lab)
Fax: +44 (0) 207 942 5867

http://www.nhm.ac.uk/research-curation/staff-directory/zoology/s-williams/in
dex.html

[Diogo Silva]

Hello Suzanne

This phenomenon of obtaining younger estimates using *BEAST was
already shown in McCormack et al. (2010) Evolution, 65:184. To my
knowledge, this is the first and only paper which compares empirical
results from BEAST and *BEAST (If I'm not entirely correct, I would
appreciate if anyone could point out additional studies). However,
given the novelty of *BEAST, it wouldn't be surprising to see a flurry
of studies with similar results in the near future. This paper also
discusses why this occurs. Shortly, the main reason is the shift from
the concatenation method (used in BEAST) to the multi-locus coalescent
approach (*BEAST), in which coalescent theory is incorporated into the
analysis when studying multiple species (basically, a shift from gene
trees to species trees). For example, the timing of gene divergence
generally predates that of speciation and thus, estimates based solely
on gene trees will tend to overestimate divergence dates. The *BEAST
method, on the other hand, uses a full probabilistic coalescent
framework, in which gene trees are embedded and jointly estimated with
the species tree, providing much more accurate and precise divergence
time estimates. Since they take this deeper gene tree coalescence into
account, estimates will be more recent. Anyway, the paper presents a
much more complete explanation.

Hope it helps

Cheers
Diogo

On Mar 24, 12:54 pm, Suzanne Williams <s.willi...@nhm.ac.uk> wrote:
> Hi Diogo,
>
> I was very interested to see that you have identified that species
> divergences seem to be younger in *BEAST analyses than BEAST analyses. We
> have found a similar pattern with our data.
>
> Do you know whether this finding has been published somewhere (so we can
> cite it)? And do you have any idea why this is occurring?
>
> Thanks for your help.
>
> Suzanne Williams
>

> http://www.nhm.ac.uk/research-curation/staff-directory/zoology/s-will...
> dex.html

[Suzanne Williams]

Hi Diogo,

Thanks very much!

Suzanne

http://www.nhm.ac.uk/research-curation/staff-directory/zoology/s-williams/in
dex.html

[pepster]

Hi,

In general, Divergence times based on the multi-species coalescent
(*BEAST, BEST) will be younger than the divergence times of the genes.
There is nothing novel about that in my opinion, it is a simple
consequence of the basic theory. Depending on population size(s),
genes can sometimes overestimate the species divergence times by a
large amount.

-Joseph

On Mar 25, 4:54 am, Suzanne Williams <s.willi...@nhm.ac.uk> wrote:
> Hi Diogo,
>

> Thanks very much!
>
> Suzanne

> http://www.nhm.ac.uk/research-curation/staff-directory/zoology/s-will...
> dex.html