no. of partitions in RAxML 7.2.8-ALPHA

[Karen]

Hi,
is the number of partitions in RAxML-7.2.8-ALPHA limited? I have a
dataset with more than 150 partitions and would like to conduct a
partitioned analysis on a superalignment (multiple genes on aa level).
If the number is limited, is there an option to change this?
Furthermore I wonder if a partitioned analysis with different models
(estimated e.g. by ProtTest) for each partition (aa level, 50-300
sites) might be useful or not - maybe they are too short for a proper
model estimation?
Thanks, Karen

[Alexis]

Dear Karen,

> is the number of partitions in RAxML-7.2.8-ALPHA limited?

yes it's limited to 128

> I have a
> dataset with more than 150 partitions and would like to conduct a
> partitioned analysis on a superalignment (multiple genes on aa level).
> If the number is limited, is there an option to change this?

I wouldn't call this an option, but there is a way to change this.

Edit the following line in source file axml.h

#define NUM_BRANCHES 128

to read

#define NUM_BRANCHES 128

then, just re-compile the program.

Don't forget to remove all object files by typing "rm *.o" first
before you re-compile.

> Furthermore I wonder if a partitioned analysis with different models
> (estimated e.g. by ProtTest) for each partition (aa level, 50-300
> sites) might be useful or not  - maybe they are too short for a proper
> model estimation?

It may make sense to do this, since your trees don't have that many
taxa.

In fact on my software page at www.exelixis-lab.org/software.html if
you scroll down
there is a perl script to do this, i.e., split up the alignment into
per-gene alignments, build a parsimony tree
for each of them, and then just compute the likelihood on the fixed
tree for all available models and keep the best one.

Cheers,

Alexis

[Karen]

Dear Alexis,
thanks a lot! (something like this I remember)
Can you specify "many taxa"? would you recommend until a certain no.
of taxa (e.g.50, 100, 1000?)
I will figure out if this is possible since I have no access to these
files (have to rely on Sys-Admins)
cheers Karen

[Alexis]

even 1000 taxa may be okay, since the model estimates are pretty
stable over different tree topologies, that is, as long
as the tree is not completely wrong, a modeltest should (hopefully)
return always the same best model over a large set of reasonable trees

Alexis

[Karen]

THANKS!!
Karen

[Karen]

sorry again a request - you wrote:
Edit the following line in source file axml.h
#define NUM_BRANCHES 128
to read
#define NUM_BRANCHES 128

-> did you mean

change the line to
#define NUM_BRANCHES X
x= no. pf partitions?

thanks
karen

[Alexis]

yes, sorry that's what I mean. By the way, you don't need to rely on
the sys-admins, you can compile a local copy
of the code.

Alexis

[Karen]

thanks alexis, i will try to figuer this out!
karen

[Karen]

Hi there,

in earlier versions with a dataset more then 128 partitions the file axml.h the no. had to be adjusted

change the line to
#define NUM_BRANCHES   X
x= no. pf partitions?

is this still the case for RAxML 8.2.X ?
And must this be exactly the no. of partitions I use or only the upper limit?

Thanks, Karen

[Alexandros Stamatakis]

that's still the case, but you only get this warning for analyses under
-M, that is, per partition branch length estimates ...

alexis

> --
> 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