Re: HKY OR GTR model select for nuclear sequence?

[Krzysztof M. Kozak]

Hi Yu,

In answer to both your posts:

1. The easiest solution is to use Modeltest to select the best fitting model.

2. Do you mean fixing the parameters obtained from Modeltest? You could do it, although it is not recommended for a Bayesian analysis. It is better to specify the substitution model and let BEAST figure out the parameters on its own.

3. In BEAUti under the MCMC tab you will find an option to generate an empty prior .xml. If you are using partitioning into codon positions you may need to edit the xml file: make sure every empty sequence is 'NNN'.

Cheers,

Chris

On Sunday, October 14, 2012 1:41:27 PM UTC+1, yi wrote:

Hello
I am a new user of Beast. When we set the site Substitution model which model should be select?   HKY OR GTR? I use the BEAST v1.7.2, can I use the model test score to fix the  Substitution model ?

Thanks,

yu

[Guy Baele]

Or you could simply do the model selection in BEAST (i.e. Bayesian model testing), instead of using Modeltest, which uses PhyML (i.e. maximum likelihood).

There's a tutorial on Bayesian model selection on the BEAST website.

I agree that fixing the parameters is not recommended. Just estimate them and make sure proper priors are used for your parameters.

Best regards,

Guy

Op maandag 15 oktober 2012 17:48:12 UTC+2 schreef Krzysztof M. Kozak het volgende:

[Jacob Berv]

I've had good success with PartitionFinder by Rob Lanfear - new software that can simultaneously analyze an entire partitioned dataset and find the optimal scheme. 

[Cuong trinh dinh]

Dear all,

Now I am using the BEAST v 2.1.3. I don´t know to select which the site substitution model, because my data fit with the GTRIG model (when i ran the Jmodeltest), but in the BEAST V2.1.3 has´t got this site substitution model (only has the JC69. GTR, HKY, TN93)? Can anybody give me a advice? 

Thanks so much!

Vào 14:41:27 UTC+2 Chủ nhật, ngày 14 tháng mười năm 2012, yi đã viết:

[Remco Bouckaert]

GTRIG = GTR + invariant sites + gamma heterogeneity.

To set this up in BEAUti 2.1.3, in the site model panel, select GTR as substitution model.
For using invariant sites, set proportion invariant to something between 0 and 1, say 0.1, and click the estimate check box next to it.
For using gamma heterogeneity  and set Gamma Category Count to 4. A shape parameter appears and click the estimate check box next to it.

Cheers,

Remco

[Matt Buys]

Sorry - had a look at the website. http://www.beast2.org/wiki/index.php/Tutorials  - which tutorial are you referring to?

[Cuong trinh dinh]

Thanks very much and dear Matt Buys i already read!

Best regards!

Vào 04:34:18 UTC+2 Thứ năm, ngày 31 tháng bảy năm 2014, Matt Buys đã viết:

[石勇]

Hi Remco,

    The best model suggested by jModeltest for my data is TPM2+G, which is not included in starBeast. 

     Should I select the GTR model and set the rate AC/AG/AT/CG/CT/GT according to the TPM2+G model and check the "estimate" box?

     Could you give me some advices? 

   Thank you very much.

Cheers,

Sandy

在 2014年7月31日星期四UTC+8上午4时13分23秒，Remco Bouckaert写道：