BEAUti can set 40 models?

[Qiyun Zhu]

Hello all,

I think I have found ways to specify 10x4=40 nucleotide substitution models by clicking in BEAUti (not by editing xml file). I'm not sure if I understand correctly.

BEAUti already has three models: HKY, GTR, TN93 (TrN).

By setting base frequencies to all equal, one gets K2P (K80) (from HKY), SYM (from GTR), TrNef (from TrN).

By unchecking operators ac, ag, at, cg, gt, one gets F81 (from GTR) and JC69 (from SYM).

By unchecking operator ag only, one gets TVM (from GTR) and TVMef (from SYM)

There leaves K3P(ef) and TIM(ef), which have to be set by editing xml file.

Are these correct? I used them in my analyses and wish to confirm I haven't done anything wrong. Thanks!

[Iv]

Hi Qiyun,
Sounds right to me, but did you get any confirmation?
Thanks!
-Iv

[Andrew Rambaut]

Dear BEAST users,

Thanks to a number of astute users who noticed some differences between runs of 1.7.2 and 1.7.3 we have discovered an issue in the latest BEAST version. Here are the details:

A flag was introduced in 1.7.3 which meant certain priors could flag that they should be evaluated first because they may return a 0 likelihood and thus short-circuit the full evaluation. Priors that set this flag are Uniform distribution and any positive distribution with an offset (values less than the offset give a 0 likelihood). This was predominantly introduced to improve performance.

Unfortunately priors with this flag set were not being included in the total density. The consequence is that uniform priors were not being respected and an prior with an offset were being ignored (this will include offset lognormal or exponential calibration priors and priors in asymmetric BSSVS phylogeography model).

Many other runs will not be effected by this issue. One solution for runs affected by this issue is to use BEAST 1.7.2 until we can release a fix. However, this option is not suitable for BSSVS discrete trait models because the above early evaluation flag was also introduced to fix an issue with this model.

Apologies for this bug. We will release a fixed version very shortly.

Andrew

___________________________________________________________________
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.

[Qiyun Zhu]

Hello lv,

I so far didn't get any confirmation :(

Best,

[Iv]

Hi,
As I understand it, by turning the operators OFF you are basically telling the program not to propose changes, thus the values should stay in their initial value and not change.  Below are two other pieces that seem to point along those same lines
-Iv

See:

http://beast.bio.ed.ac.uk/FAQ#How_to_make_use_of_other_substitution_models_not_given_in_Beauti.3F_Specifically.2C_Jukes-Cantor_model.3F

Hi Andrew,

The F81 model is almost the same as the simple Jukes-Cantor model,
except it allows unequal base frequencies. So you can simply select
the GTR model in BEAUti and turn off the operators for all of the rate
parameters.

Cheers,
Simon

- hide quoted text -

On Nov 11, 10:45 am, Andy <andrew.gotts...@gmail.com> wrote:
> Hi all,
>
> Does anyone have the xml code to implement the F81 mutation model in
> BEAST? I did a search in this group and I did not find the code.
>
> Thanks,
>
> Andrew Gottscho

[Qiyun Zhu]

Hello lv,

Yes that's also what I understood. I tried my way on some analysis (particularly TVM model). It seems things are working fine.

Maybe there are ways to specify other models. I wish I can learn how.

Best,

[Andrew Rambaut]

Dear BEAST users,

We have just released BEAST v1.7.4 which fixes a critical bug with the previous version (see the details in my previous posting, below). Essentially, we would recommend re-running any analysis that included user-specified Uniform priors, or other distributions such as lognormal or exponential priors with non-zero offsets. Also discrete trait/phylogeography models where BSSVS and asymmetric rate matrices were being used.

You don't have to recreate the XML files as BEAUti is largely unchanged (one additional bug has been fixed but this would simply have prevented XMLs being generated).

This bug was introduced in v1.7.3 so runs with previous versions should also be OK.

The new downloads are in the usual place:

http://code.google.com/p/beast-mcmc/downloads/list

Thanks and apologies,
Andrew

> --
> 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.

[石勇]

Hi, Qiyun

 where is the operator? Do you mean Beast or starBeast?

Cheers,

Sandy

在 2012年8月12日星期日UTC+8上午11时45分15秒，Qiyun Zhu写道：

[Qiyun Zhu]

Hello Yong,

I am referring to BEAST, but I think it applies to *BEAST too. There is an "Opterator" panel in BEAUti.

[Sonal Mundhra]

Hi..
I am a new BEAST user. I wish to ask a few things.

I am using the substitution model TN93 with site heterogeneity model gamma plus invariant sites and no of gamma categories 5. I tried using base frequency as empirical but when I ran the programme, it showed an error that "The initial model is invalid because state has a zero probability." I am using ver 1.7.5.

The other parameters were as follows:
Clock: exponential relaxed, estimate checked
tree: bayesian skyline, number 10, piecewise -constant
Priors: uced mean set to exponential with initial value, mean and offset value set at 1.0, 1.0 and 0.0 respectively
other priors default
MCMC chain length 300000000.

But i am getting that error as mentioned above only with base frequency set to empirical not when it is set to estimated.

Any reasons why it is happening and any other parameter or prior I need to modify.

Thanks.