XPLOR-NIH on Biowulf
XPLOR-NIH is a structure determination program which builds on the X-PLOR
v 3.851 program, including additional tools developed at the NIH.
These tools
include functionality for the following:
- 3J couplings
- 1J couplings
- 13C shifts
- 1H shifts
- T1/T2
- dipolar couplings
- radius of gyration
- CSA
- conformational database torsion angle potentials
- database base-base positioning potentials for DNA
- interface to the NMR graphics package
VMD-XPLOR,
downloadable separately.
- embedded Python and TCL interpreters.
The advantage of
running Xplor-NIH on Biowulf would be to spawn a large number of
independent refinement jobs which would run on multiple Biowulf nodes.
Note that each run is not parallel; the speedup is produced by
the swarm of independent jobs. Thus it is ideally suited for refining
multiple NMR structure models. Xplor-NIH is based on Xplor 3.851.
XPLOR was developed
at Yale University, and the NIH extensions were developed by G. Marius
Clore, John Kuszewski, Charles D. Schwieters, and Nico Tjandra.
XPLOR-NIH web site,
including tools for use at
NIH
Xplor Manual
Running Xplor-NIH on Biowulf
Xplor-NIH requires the usual set of Xplor input files -- coordinates,
xplor input files, topology/parameter files. The 'standard' files are
in /usr/local/xplor-nih/.
To submit an Xplor-NIH batch job on Biowulf, you need a PBS batch
script.
Sample script, modified from Charles Schwieters.
#/bin/sh
#
# This is a PBS script to run parallel XPLOR-NIH jobs on Biowulf
# CDS - 12/16/02
#
#
# PBS stuff
#
#PBS -N xplor-nih
#PBS -S /bin/sh
#PBS -m be
#PBS -k oe
#
# cd to the 'working' directory, from where the job is being submitted.
# This can also be set to an explicit pathname.
#
cd $PBS_O_WORKDIR
#
# uncomment if the script is a python script
#
#PY=-py
#
# name of XPLOR script
#
XPLOR_SCRIPT=test.inp
#
# name of output file (output will appear in script.out.#)
# where # is 0-(num processors)
#
LOGFILE=script.out
#
#location of xplor
#
XPLOR=/usr/local/xplor-nih/bin/xplor
#
#list of nodes on which to run:
# this to run two jobs on each node. Do not change this.
#
XPLOR_MACHINES=xplor_machines.$PBS_JOBCOOKIE
cat $PBS_NODEFILE $PBS_NODEFILE > $XPLOR_MACHINES
#
# run xplor
#
env XPLOR_RSH="rsh -n" \
$XPLOR $PY -parallel -o $LOGFILE -machines $XPLOR_MACHINES \
< $XPLOR_SCRIPT
#
# clean up
#
rm $XPLOR_MACHINES
|
Submit this job with a command like:
qsub -l nodes=8 scriptname
Monitoring your job
\As with all jobs on Biowulf, the individual job
or the entire system can be monitored with a variety of web-based tools.
On the right is an
image of the monitor for all jobs belonging to one user.
See the Biowulf user
guide for a description of these tools.
Biowulf Cluster monitors
|
 |
Benchmarks
Xplor-NIH test suite.
| Node type | Time
o2200 2.2 GHz AMD Opteron 248 | 49 seconds
| o2000 2.0 GHz AMD Opteron 246 | 54 seconds
| p2800 2.8 GHz Intel Xeon | 109 seconds
| p1800 1.8 GHz AMD Athlon | 120 seconds
| p1400 1.4 GHz AMD Athlon | 141 seconds
| |
|