[maker-devel] altest without MPI?

[Carson Holt]

It's best to use the cegma results with the cegma2zff script to generate a
training set for SNAP.  Then don't use the cegma proteins.  If you can get
proteins from a related species with an annotated genome, it will be
better than altest option for a different species.  This is because altest
is aligned via tbalstx which is 3-4 time slower than protein alignments.
Also they will rarely be good enough to produce many est2genome models
(best to only use them if you have nothing else).

The cpus= option is a blast parameter for specifying how many cpus to give
to each blast job.  It is not an MPI parameter.  The number of cpus for
MPI is specified using the -n option from mpiexec and not in the maker
control files.

You don't have to use MPI.  You can also split your contigs up into
separate jobs and run MAKER multiple times.  Use the fasta_tool script
that comes with MAKER to split your input file up.

Let us know if you come across anything you have more questions on.

Thanks,
Carson



On 13-06-17 1:20 PM, "Jacqueline R M Doyle" <jmdoyle at purdue.edu> wrote:

>Hi!
>
>I am beginning my first MAKER annotation and had a quick question.  I am
>currently planning on following the ³Training ab initio Gene Predictors²
>section of the MAKER 2012 tutorial.  For my species of interest, I have
>784290 scaffolds in which 80% are greater than 100 kb.  I have EST data
>from a closely related species and was also going to use the core cegma
>protein sequences.  With this in mind, I made the following changes to my
>maker_opts file:
>
>genome=scaffolds.fasta
>altest=Trinity.fasta
>protein=cegma.fa
>est2genome=1
>cpus=48
>
>My primary concern is that this is going to take a long time to run with
>altest, even with the extra cpus for BLAST.  The software was not
>originally installed on our computer cluster with MPICH2, but I may be
>able to talk our computer guys into reinstalling if the situation is
>going to be completely untenable without MPI.  I guess my question is, is
>there any point in trying to run the above without MPI?  Is there a good
>way to monitor the progress of such a run if I was to give it a shot?
>
>Thanks for your help with this!
>
>Jackie
>
>_______________________________________________
>maker-devel mailing list
>maker-devel at box290.bluehost.com
>http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org