Convergence criterion (-D) in rapid BS+ML analysis (-f a)

[Alexander Esin]

Hi there,

I'm running RAxML on a very heterogeneous protein dataset in terms of number of taxa, and I am practically limited by a maximum walltime on the server I use. For a subset of trees, I would still like to acquire BS replicates - but would like to also turn on the ML search convergence criterion to speed up the ML step of the analysis. My question is, does the -D argument work with -f a, for example like in the call below?

raxml -f a -D -m PROTCATAUTO -p 1234 -x 10001 -N 100 -T 4 -s alignment -n name

If not, would you suggest running the rapid bootstraps first (via -x), then building the ML trees (-D), and then adding the BS values onto the tree in a third step?

Thanks!

Alex

[Alexandros Stamatakis]

Dear Alex,

> I'm running RAxML on a very heterogeneous protein dataset in terms of
> number of taxa, and I am practically limited by a maximum walltime on
> the server I use. For a subset of trees, I would still like to acquire
> BS replicates - but would like to also turn on the ML search convergence
> criterion to speed up the ML step of the analysis. My question is, does
> the -D argument work with -f a, for example like in the call below?
>
> raxml -f a -D -m PROTCATAUTO -p 1234 -x 10001 -N 100 -T 4 -s alignment
> -n name

I just looked at the source code and in fact -D does not have an effect
with -f a, I'll update the documentation accordingly.

> If not, would you suggest running the rapid bootstraps first (via -x),
> then building the ML trees (-D), and then adding the BS values onto the
> tree in a third step?

That is a possibility, alternatively you could use ExaML if you are
limited by wall-time, ExaML has a checkpoint and restart capability.

Alexis

>
> Thanks!
>
> Alex
>
> --
> 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.

--
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

[Alexander Esin]

Dear Alexis,

Thank you for the swift reply - my benchmarking did not show a difference, which is why I asked!

I will look into ExaML - would it be compatible for use on a ccNUMA shared memory system?

Best,

Alex

[Alexandros Stamatakis]

Dear Alex,

> Thank you for the swift reply - my benchmarking did not show a
> difference, which is why I asked!

:-)

> I will look into ExaML - would it be compatible for use on a ccNUMA
> shared memory system?

Yes, you can either run MPI on a ccNUMA, that is, one MPI process per
core or you can run the new hybrid MPI/OpenMP version of ExaML, which
option is faster actually depends on the system, we have included some
performance plots in the latest ExaML manual. Also, ExaML exclusivley
parallelizes likelihood calculations over sites, thus if your alignment
is short in length, the degree of parallelism you can exploit will be small.

Alexis

> > an email to raxml+un...@googlegroups.com <javascript:>
> > <mailto:raxml+un...@googlegroups.com <javascript:>>.

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

> <https://groups.google.com/d/optout>.

>
> --
> 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 <http://www.exelixis-lab.org>

[Alexandros Stamatakis]

RAxML v 8.1.21 (just released) will now exit with an error if you
specify -D in conjunction with -x or -x and -f a,

alexis

> > an email to raxml+un...@googlegroups.com <javascript:>
> > <mailto:raxml+un...@googlegroups.com <javascript:>>.

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

> <https://groups.google.com/d/optout>.

>
> --
> 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 <http://www.exelixis-lab.org>