Setting MRCA for 25 partitions

77 views
Skip to first unread message

Jamie Thompson

unread,
Jun 2, 2020, 2:20:40 PM6/2/20
to beast-users
Hiya,

I'm running bmodeltest with c.300 taxa and 25 gene partitions.  When setting MRCA in the priors tab of BEAUTI, do I just need to set my 15 calibrations for one locus, or repeat for every locus?  The latter will take a long time!

Best,
Jamie

Remco Bouckaert

unread,
Jun 2, 2020, 7:10:14 PM6/2/20
to beast...@googlegroups.com
Hi Jamie,

If you have 25 gene partitions, you probably want to do one of these two:

o link the tree for all partitions: this has the effect of having a single gene tree and the alignment will be concatenated. This is appropriate if the genes all have a common history (e.g. all loci are mitochondrial). In this case, the MRCA prior needs to be defined on the single tree only.

o if (some of) the loc do not share a common history, use a multi species coalescent analysis (*BEAST, or — more efficiently — with the StarBeast2 package). The loci will then have their individual gene trees (assuming they all have their own history, otherwise you may need to link trees for some of them where appropriate) fitting in a single species tree, and you define a single MRCA prior on the species tree.

Cheers,

Remco


--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beast-users/cf447ed4-13d4-4d42-bcd6-8ebf232d591d%40googlegroups.com.

Carlo Pacioni

unread,
Jun 2, 2020, 9:16:52 PM6/2/20
to beast...@googlegroups.com
Just to complement Remco's reply, if you are doing demographic analyses where you are using, for example, EBSP then indeed you have to keep the partitions that are independent as unlinked trees, and have to apply a calibration to each of those. I find it easier to put the calibration for one tree in BEAUti, save the XML file, and then copy and paste the calibration block as many time as you need and simply change 'YourTree' in the attribute tree="@Tree.t:YourTree" with the names of your partitions (you can copy and paste them from other parts in the XML). You will see that the distribution that you use in the calibration has one or more attributes that look like  id="RealParameter.9". You also have to change the number after the dot. This is a random number, so you have to change it to anything that has not been used somewhere else in the xml.

If you only have to do this once, that shouldn't be a problem, but if you have to set up several analyses I think the R package "beautier" may come in handy for this kind of things.

cheers,
carlo

Cheers,
carlo



Jamie Thompson

unread,
Jun 3, 2020, 3:27:37 PM6/3/20
to beast-users
Hi Remco,

Thanks for the advice.  My alignment is a supermatrix of published data with mitochondrial and nuclear sequences, and even the mitochondrial sequences are unlikely to come from the same individual.  Additionally, not all taxa are represented by every gene, some genes have 230 taxa and some 30, with lots of missing data.  With Starbeast, can you supply only one mrca prior, am I reading his correctly?

Best wishes and many thanks,
Jamie

On Wednesday, June 3, 2020 at 12:10:14 AM UTC+1, Remco Bouckaert wrote:
Hi Jamie,

If you have 25 gene partitions, you probably want to do one of these two:

o link the tree for all partitions: this has the effect of having a single gene tree and the alignment will be concatenated. This is appropriate if the genes all have a common history (e.g. all loci are mitochondrial). In this case, the MRCA prior needs to be defined on the single tree only.

o if (some of) the loc do not share a common history, use a multi species coalescent analysis (*BEAST, or — more efficiently — with the StarBeast2 package). The loci will then have their individual gene trees (assuming they all have their own history, otherwise you may need to link trees for some of them where appropriate) fitting in a single species tree, and you define a single MRCA prior on the species tree.

Cheers,

Remco

On 3/06/2020, at 2:41 AM, Jamie Thompson <thompso...@gmail.com> wrote:

Hiya,

I'm running bmodeltest with c.300 taxa and 25 gene partitions.  When setting MRCA in the priors tab of BEAUTI, do I just need to set my 15 calibrations for one locus, or repeat for every locus?  The latter will take a long time!

Best,
Jamie

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages