New to ABySS - Proper execution and parameters for large assembly

[JPummil]

Greetings!

First, a bit of history. I am the manager for the Arkansas High
Performance Computing Center (hpc.uark.edu) and am assisting our first
Bio users on the HPC cluster that services our researchers. The ABySS
code built with no problem on the system and is globally accessible to
all users:

$ ./configure --prefix=/share/apps/abyss/abyss-1.3.3/ --with-boost=/
share/apps
/boost/include/ --with-mpi=/share/apps/mpi/openmpi-1.5.1/intel-12.0.0/
include --
disable-openmp CC=mpicc CXX=mpicxx CPPFLAGS=-I/share/apps/
sparsehash-2.0.2/inclu
de/

We have also created a module file that allows perspective users to
load the path to all of the ABySS binaries into their user
environment.

Our systems use Torque and Maui for job scheduling and resource
management.

Now, to the problem we want to run using ABySS...

So, first group (4 files) is single reads while second group (2 files)
is paired-end data.

stargate:drhoads:/fasttmp/drhoads/Illumina$ ls -l
total 104701152
-rw-r--r-- 1 drhoads drhoads 28789945446 Jun 6 2011
s_1_1_sequence.txt
-rw-r--r-- 1 drhoads drhoads 28789945446 Jun 6 2011
s_1_2_sequence.txt
-rw-r--r-- 1 drhoads drhoads 24816955400 Jun 6 2011
s_2_1_sequence.txt
-rw-r--r-- 1 drhoads drhoads 24816955400 Jun 6 2011
s_2_2_sequence.txt
stargate:drhoads:/fasttmp/drhoads/Illumina$ cd ../IlluminaPE/
stargate:drhoads:/fasttmp/drhoads/IlluminaPE$ ls -l
total 42937484
-rw-r--r-- 1 drhoads drhoads 21983941866 Mar 18 09:08
Rattle_Snake__RS_212M__L3_GGCTAC_L003_R1_001.fastq
-rw-r--r-- 1 drhoads drhoads 21983941866 Mar 18 11:50
Rattle_Snake__RS_212M__L3_GGCTAC_L003_R2_001.fastq

I haven't added the scheduler directives yet, but I was hoping for
some advice on the attempted submission line:

abyss-pe k=64 name=crotalus lib='IlluminaPE'
'IlluminaPE='Rattle_Snake__RS_212M__L3_GGCTAC_L003_R1_001.fastq
Rattle_Snake__RS_212M__L3_GGCTAC_L003_R2_001.fastq'
se='s_1_1_sequence.txt s_1_2_sequence.txt s_2_1_sequence.txt
s_2_2_sequence.txt'

My goal here is to get the HPC system details worked out and then turn
things back to the Bio researchers for fine tuning of the more
intricate features of the code. I have NOT attempted to run this yet,
but was optimistic that I could get a few pointers from the community
ahead of time?

Best,

Jeff Pummill
AHPCC

[Shaun Jackman]

Hi Jeff,

Your command line looks good. Give it a run!

Disabling OpenMP is not ideal, but it will work. The stage that maps the reads back to the assembly is a time-consuming step and is parallelized using OpenMP.

Cheers,
Shaun