Asymmetric substitution model for phylogeography

[geo]

dear all
I have some doubt about phylogeographic analysis using asymmetric substitution model (in association with BSSVS).
how can I decide how to choose symmetric or asymmetric substitution model? only on epidemiological knowledges  or can I use method like path sampling or stepping stone??
it is still possible to use sòread to define well supported migration rate?? if it is possible, what value of poisson priori mean and offset should I set??
thank you for any help!!
Giovanni

[alex baumel]

Good question !

alex

[Andrew Rambaut]

Dear Giovanni,

I can see no reason why you shouldn't use path-sampling to look at the relative fit of the asymmetric and symmetric models to your data.

SPREAD can visualize both models.

The default Poisson priors on non-zero rates generated by BEAUti differ between the two models. For the symmetric model the suggested (I hesitate to say 'recommended') prior is a Poisson with mean = 0.693 (log2) and the offset = K - 1 (where K is the number of states). This has the effect of putting 50% probability on getting K-1 rates which is the minimum necessary to connect all the states.

For the asymmetric model it is a simpler Poisson with mean of K-1 (offset 0).


Andrew

> --
> 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 post to this group, send email to beast...@googlegroups.com.
> Visit this group at http://groups.google.com/group/beast-users?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

___________________________________________________________________
Andrew Rambaut
Institute of Evolutionary Biology University of Edinburgh
Ashworth Laboratories Edinburgh EH9 3JT
EMAIL - a.ra...@ed.ac.uk TEL - +44 131 6508624


--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

[geo]

Thank you very much!!!

[Jason Tokayer]

I have a few additional questions regarding the xml file creation for symmetric vs asymmetric substitution models. It seems to me that when selecting an asymmetric model BEAUTi changes the xml file in 3 ways:

(1) In the generalSubstitutionModel definition, randomizeIndicator="false" appears when using the asymmetric option

(2) In the rates attribute inside the generalSubstitutionModel definition the dimension is doubled

(3) In the ancestralTreeLikelihood definition for the locations, an additional frequencyModel is added and titled "locations.root.frequencyModel".

The reason for (2) is obvious to me, but I don't understand if/why (1) and (3) are necessary. Can someone provide an explanation?

Thanks,

Jason

[Philippe Lemey]

Dear Jason,

As for 1), the option randomizeIndicator="false" simply became visible in the xml, but the default is false anyway. Only for BSSVS, it may be useful to have a starting state where the rate indicators are randomly set to 0 or 1, which can be achieved be setting randomizeIndicator="true" 

Part of the answer for 3) is formulated in this thread: 

https://groups.google.com/forum/#!searchin/beast-users/gianni/beast-users/MyCYPCnqhDQ/5GAY7u9xnkYJ

best,

Philippe

--
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 post to this group, send email to beast...@googlegroups.com.

Visit this group at http://groups.google.com/group/beast-users.

[Jason Tokayer]

Thanks for the link, Philippe. Although the frequency model has id "root.frequencyModel", I don't see how this is actually connected to the root. I assume this is built into the BEAST code so that any asymmetric CTMC model converges to the root frequencies - do you (or anyone else) have a reference publication that I can look at to help me understand the nonreversible model (wrt phylogenetics) a bit better? 

Also, I am looking at the xml file for the GLM substitution model from Nuno Faria's paper. In this case, the <rootFrequencies> tag is used for the frequency model. However, the paper makes no mention of the theory behind this (in fact there is no mention to a frequency model at all). Might you (or anyone else) have a reference publication for this?