RAxML: how to avoid final gamma optimisation using -V

[Cymon Cox]

Hi Folks, I hope someone can help me out here...

I want to run 10 ML search replicates under a GTR + ML estimated composition (X)

The -h says:

 -V      Disable rate heterogeneity among sites model and use one without rate heterogeneity instead.
              Only works if you specify the CAT model of rate heterogeneity.


By which I interpret the following should work (using RAxML build from github today):

raxmlHPC-SSE3-8.2.4 -s <data file> -m GTRCATX -V -f d -n run1 -p 1234 -N 10

-f d ; -f a ; -f o - but all optimise the gamma at the end of searches, and gives this warning:

Conducting final model optimizations on all 10 trees under GAMMA-based models ....
 
WARNING the alpha parameter with a value of 12.993029 estimated by RAxML for partition number 0 with the name "No Name Provided"
is larger than 10.000000. You should do a model test and confirm that you actually need to incorporate a model of rate heterogeneity!
You can run inferences with a plain substitution model (without rate heterogeneity) by specifyng the CAT model and the "-V" option

So, what combination of options allows me to avoid the final optimisation?

Thanks, Cymon

[Alexey Kozlov]

Hi Cymon,

please try "-F" option, it should disable the final optimization under GAMMA.

But AFAIK this warning is only there to prevent the waste of resources (GTRCAT is usually much faster). So the results
should be still valid even if you use GAMMA on the dataset without rate heterogeneity. Finally, please note that your
alpha value (12.99) is just slightly higher than the threshold.

Alexey

On 17.02.2016 17:20, 'Cymon Cox' via raxml wrote:
> Hi Folks, I hope someone can help me out here...
>
> I want to run 10 ML search replicates under a GTR + ML estimated composition (X)
>
> The -h says:
>
> -V Disable rate heterogeneity among sites model and use one without rate heterogeneity instead.
> Only works if you specify the CAT model of rate heterogeneity.
>
>
> By which I interpret the following should work (using RAxML build from github today):
>
> raxmlHPC-SSE3-8.2.4 -s <data file> -m GTRCATX -V -f d -n run1 -p 1234 -N 10
>

> -f d ; -f a ; -f o - *but all optimise the gamma at the end of searches*, and gives this warning:
>
> /Conducting final model optimizations on all 10 trees under GAMMA-based models ..../
> //
> /WARNING the alpha parameter with a value of 12.993029 estimated by RAxML for partition number 0 with the name "No
> Name Provided"/
> /is larger than 10.000000. You should do a model test and confirm that you actually need to incorporate a model of
> rate heterogeneity!/
> /You can run inferences with a plain substitution model (without rate heterogeneity) by specifyng the CAT model and
> the "-V" option/

>
>
> So, what combination of options allows me to avoid the final optimisation?
>
> Thanks, Cymon
>

> --
> You received this message because you are subscribed to the Google Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

[Cymon Cox]

On 17 February 2016 at 17:53, Alexey Kozlov <alexei...@gmail.com> wrote:

Hi Cymon,

please try "-F" option, it should disable the final optimization under GAMMA.

But AFAIK this warning is only there to prevent the waste of resources (GTRCAT is usually much faster). So the results should be still valid even if you use GAMMA on the dataset without rate heterogeneity. Finally, please note that your alpha value (12.99) is just slightly higher than the threshold.

Alexey

 

Thanks Alexey, much appreciated... (it's right there in the -h, but I just couldn't see it!)

Cheers, Cymon

 

--
You received this message because you are subscribed to a topic in the Google Groups "raxml" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/raxml/M29ocy0_W8o/unsubscribe.
To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

____________________________________________________________________

Cymon J. Cox


FCT Investigador - Coordinating Researcher
Plant Systematics and Bioinformatics Research Group (PSB)
Centro de Ciencias do Mar (CCMAR) - CIMAR-Lab. Assoc.

Mailing address:
CCMAR - Centro de Ciencias do Mar,
Universidade do Algarve
Campus de Gambelas
Edif. 7
8005-139 Faro
Portugal

Phone: +351 289800051 ext 7380
Fax: +351 289800051
Email: cymo...@googlemail.com
HomePage : http://www.ccmar.ualg.pt/home/index.php?id=202
GPG: Public key on keyserver.ubuntu.com
Scopus ID: 7402112716
ResearcherID: D-1303-2012

Orcid ID: 0000-0002-4927-979X

[Cymon Cox]

Hi Alexey,

On 17 February 2016 at 17:58, Cymon Cox <cymo...@googlemail.com> wrote:

On 17.02.2016 17:20, 'Cymon Cox' via raxml wrote:

Hi Folks, I hope someone can help me out here...

I want to run 10 ML search replicates under a GTR + ML estimated composition (X)

The -h says:

  -V      Disable rate heterogeneity among sites model and use one without rate heterogeneity instead.
               Only works if you specify the CAT model of rate heterogeneity.

On 17 February 2016 at 17:53, Alexey Kozlov <alexei...@gmail.com> wrote:

Hi Cymon,

please try "-F" option, it should disable the final optimization under GAMMA.

But AFAIK this warning is only there to prevent the waste of resources (GTRCAT is usually much faster). So the results should be still valid even if you use GAMMA on the dataset without rate heterogeneity. Finally, please note that your alpha value (12.99) is just slightly higher than the threshold.

Alexey

 

I guess I'm just realising what's going on here:

You can only use the -V option when using the CAT model. But when using, say -m GTRCATX -V, it optimises the trees with GAMMA at the end on all X reps and identifies the best tree. If you include -F it doesnt optimise anything in the end, and you have X number of reps/trees without branch lengths and with likelihoods (which are meaningless) that are not able to distinguish the best tree among the reps (or are they comparable among reps?).

So effectively, it is not possible to do 10 random starting tree replications under GTRX and end up with one best tree with optimal branch lengths. Is that right?

Regards, Cymon

[Alexandros Stamatakis]

yes, but you should be able to take all those 10 trees and optimize them
with -m GTRCATX -V using -f e,

alexis

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
Adjunct Professor, Dept. of Ecology and Evolutionary Biology, University
of Arizona at Tucson

www.exelixis-lab.org

[Cymon Cox]

Thanks Alexis. I assume I can just ignore the references to GAMMA in the output.

Regards, Cymon

raxmlHPC-SSE3-8.2.4 -m GTRCATX -V -f e -s ../<data> -t RAxML_result.run1.RUN.0 -n run0


Model parameters (binary file format) written to: <path>/RAxML_binaryModelParameters.run0
0.127152 -1437.575932


Overall Time for Tree Evaluation 0.127373
Final GAMMA  likelihood: -1437.575932

Number of free parameters for AIC-TEST(BR-LEN): 122
Number of free parameters for AIC-TEST(NO-BR-LEN): 9


Model Parameters of Partition 0, Name: No Name Provided, Type of Data: DNA
alpha: 1.000000
Tree-Length: 0.036285
rate A <-> C: 0.000100
rate A <-> G: 1.569813
rate A <-> T: 0.339418
rate C <-> G: 0.174224
rate C <-> T: 2.342306
rate G <-> T: 1.000000

freq pi(A): 0.217643
freq pi(C): 0.273285
freq pi(G): 0.324206
freq pi(T): 0.184866

[Alexandros Stamatakis]

yes you can, this is only a bug in the printout,

alexis

>> <http://orcid.org/0000-0002-4927-979X>