Frodo: asdf

2024-05-02T18:29:59Z

asdf

← Older revision		Revision as of 18:29, 2 May 2024
Line 142:		Line 142:

	Note that as written, the first script takes a list of ''model files'' as input (i.e. "model1.pdb", etc) and the second takes a list of ''directory names'', which are by default the names of the models ("model1", etc). This is because you may want to run the analysis over only a subset of models, or over an ensemble generated by a different method, etc.		Note that as written, the first script takes a list of ''model files'' as input (i.e. "model1.pdb", etc) and the second takes a list of ''directory names'', which are by default the names of the models ("model1", etc). This is because you may want to run the analysis over only a subset of models, or over an ensemble generated by a different method, etc.

			[[Category:DOCK]]
			[[Category:Tutorials]]

Kstafford: model ensemble scripts, since this came up today

2017-06-02T20:50:36Z

model ensemble scripts, since this came up today

New page

This is a really simple script to facilitate automated setup and docking for an ensemble of models. It is intended to be used to screen a set of models for which performs best at enriching known ligands from a relatively small set of decoys/inactives (e.g. the Roth lab collection). It submits a series of jobs for each model in an input list and then submits a predefined docking job for each model after it has been prepared using blastermaster. This means that it requires each of the nodes in the first set of jobs (blastermaster, etc.) to in turn act as a submit host for the second set (docking). It should ''not'' be used for large screens.

Copied from /nfs/home/kstafford/scripts/model_ensemble_setup.py:

<pre>
#! /usr/bin/env python

import os, sys

def model_ensemble_setup(modellist,xtal,dirstring):
f = open(modellist)
g = f.readlines()
f.close()

pwd = os.getcwd()
dirlist = []

for i in range(0,len(g)):
m = g[i].strip().split('.pdb')[0]
print m
if not os.path.exists(m):
os.mkdir(m)

os.chdir(m)
os.system('ln -s ../%s m.pdb' % g[i].strip())
os.system('ln -s %s xtal-lig.pdb' % xtal.strip())

# simple, obvious way to build dirlist
dirlist.append(m+'\n')

os.chdir(pwd)

f = open('dirlist','w+')
f.writelines(dirlist)
f.close()

runtext = '''#$ -S /bin/csh
#$ -cwd
#$ -j yes
#$ -o test.out
#$ -e test.err
#$ -q all.q

echo "Starting dock in directory: $PWD"
set dirarray=`cat dirlist`
set pth=$dirarray[$SGE_TASK_ID]
cd $pth
echo $pth

$DOCKBASE/bin/blastermaster -v
$DOCKBASE/bin/setup_db2_lots 20 d "%s"
$DOCKBASE/bin/submit_dock
''' % dirstring

f = open('run_all_models.csh','w+')
f.write(runtext)
f.close()

# os.system('qsub -t 1-%d run_all_models.csh' % len(g))

return

if __name__ == "__main__":
if len(sys.argv) != 4:
print "Usage: [file containing a list of models] [full path to xtal-lig] [directory search pa
th to db2 files]"
else:
model_ensemble_setup(sys.argv[1],sys.argv[2],sys.argv[3])
sys.exit()
</pre>

After the docking is done, you can parallelize the usual analysis steps like so:

Copied from /nfs/home/kstafford/scripts/run_analysis.py:

<pre>
#! /usr/bin/env python

import os,sys

def run_analysis(dirlist,liglist,decoylist):

if not os.path.exists(liglist):
print "Ligand ID file %s doesn't exist" % liglist
return
if not os.path.exists(decoylist):
print "Decoy ID file %s doesn't exist" % decoylist
return
if not os.path.exists(dirlist):
print "Directory list %s doesn't exist" % dirlist

submit_text = '''#$ -S /bin/csh
#$ -cwd
#$ -j yes
#$ -o test.out
#$ -e test.err
#$ -q all.q

echo "Starting dock in directory: $PWD"
set dirarray=`cat %s`
set pth=$dirarray[$SGE_TASK_ID]
cd $pth
echo $pth

$DOCKBASE/analysis/extract_all.py
$DOCKBASE/analysis/enrich.py -l %s -d %s
$DOCKBASE/analysis/getposes.py

echo "started task"
echo $SGE_TASK_ID

exit $status
''' % (dirlist,liglist,decoylist)

f = open(dirlist)
dirs = f.readlines()
f.close()
ndirs = len(dirs)

f = open('run_analysis.csh','w+')
print >>f, submit_text
f.close()

os.system('qsub -t 1-%d run_analysis.csh' % ndirs)

return

if __name__ == "__main__":
if len(sys.argv) != 4:
print "Usage: [directory list] [ligand ID list] [decoy ID list]"
else:
run_analysis(sys.argv[1],sys.argv[2],sys.argv[3])
sys.exit()

</pre>

Note that as written, the first script takes a list of ''model files'' as input (i.e. "model1.pdb", etc) and the second takes a list of ''directory names'', which are by default the names of the models ("model1", etc). This is because you may want to run the analysis over only a subset of models, or over an ensemble generated by a different method, etc.

Model ensemble setup - Revision history

Frodo: asdf

Kstafford: model ensemble scripts, since this came up today