def run_probe( pdb_filename , variants , probe_output_filename = '' , run = True ):
"""
Runs PROBE on <pdb_filename> on the positions found among <variants>
using the default options in PROBE_OPTIONS and writes the output to
<probe_output_filename> (also returns this output filename)
"""
if not probe_output_filename:
probe_output_filename = os.path.abspath( pdb_filename ).rstrip( '.pdb' ) + '.probe_out'
# get the unique variant positions
positions = list( set( [i[1:-1] for i in variants] ) )
positions.sort()
# generate the commands to run
# command = '#!/bin/sh\nrm ' + probe_output_filename + '\ntouch ' + probe_output_filename + '\n'
command = 'rm ' + probe_output_filename + ';touch ' + probe_output_filename + ';'
# delete any prior copy since we will append to it
for i in positions:
probe_options = {}
probe_options.update( PROBE_OPTIONS )
probe_options['out'] = pdb_filename
probe_options['Q'] = str( i )
command += create_executable_str( PATH_TO_PROBE , [] , probe_options , probe_output_filename , append = True ) +';'#'\n'
# run PROBE, store the output
if run:
run_local_commandline( command )
return probe_output_filename , positions
else:
# the command, well, get positions etc. too
return command , probe_output_filename , positions
def run_psiblast(sequence_filename, run=True):
"""
Runs PSIBLAST on <sequence_filename> using the default options in
PSIBLAST_OPTIONS and returns the relevant output file: "out_ascii_pssm"
"""
root_filename = os.path.abspath(sequence_filename).rstrip('.fa')
# collect the options, set the input, derive the output filenames
psiblast_options = {}
psiblast_options.update(PSIBLAST_OPTIONS)
psiblast_options['query'] = sequence_filename
for i in psiblast_options.keys():
if '__call__' in dir(psiblast_options[i]):
psiblast_options[i] = psiblast_options[i](root_filename)
for i in psiblast_options.keys():
if isinstance(psiblast_options[i], str) and os.path.isfile(
psiblast_options[i]):
psiblast_options[i] = os.path.abspath(psiblast_options[i])
command = create_executable_str(PATH_TO_PSIBLAST,
args=[],
options=psiblast_options)
if run:
run_local_commandline(command)
# the only output we need
return psiblast_options['out_ascii_pssm']
else:
# just send the command
return command, psiblast_options['out_ascii_pssm']
def run_rosetta_relax_local( pdb_filename , extra_options = {} , run = True ):
root_filename = os.path.abspath( pdb_filename ).replace( '.pdb' , '' )
# collect the options, set the input, derive the output filenames
relax_options = {}
relax_options.update( ROSETTA_RELAX_OPTIONS )
relax_options.update( extra_options )
relax_options['s'] = pdb_filename
relax_options['native'] = pdb_filename # required to get gdtmm scores
for i in relax_options.keys():
if '__call__' in dir( relax_options[i] ):
relax_options[i] = relax_options[i]( root_filename )
for i in relax_options.keys():
if isinstance( relax_options[i] , str ) and os.path.isfile( relax_options[i] ):
relax_options[i] = os.path.abspath( relax_options[i] )
# ...weird Rosetta append behavior...
if os.path.isfile( relax_options['out:file:silent'] ):
os.remove( relax_options['out:file:silent'] )
if os.path.isfile( relax_options['out:file:scorefile'] ):
os.remove( relax_options['out:file:scorefile'] )
command = create_executable_str( PATH_TO_ROSETTA_RELAX , args = [] , options = relax_options )
if run:
run_local_commandline( command )
# the only output we need
return relax_options['out:file:silent']
else:
return command , relax_options['out:file:silent']
def run_rosetta_relax_local(pdb_filename, extra_options={}, run=True):
root_filename = os.path.abspath(pdb_filename).replace('.pdb', '')
# collect the options, set the input, derive the output filenames
relax_options = {}
relax_options.update(ROSETTA_RELAX_OPTIONS)
relax_options.update(extra_options)
relax_options['s'] = pdb_filename
relax_options['native'] = pdb_filename # required to get gdtmm scores
for i in relax_options.keys():
if '__call__' in dir(relax_options[i]):
relax_options[i] = relax_options[i](root_filename)
for i in relax_options.keys():
if isinstance(relax_options[i], str) and os.path.isfile(
relax_options[i]):
relax_options[i] = os.path.abspath(relax_options[i])
# ...weird Rosetta append behavior...
if os.path.isfile(relax_options['out:file:silent']):
os.remove(relax_options['out:file:silent'])
if os.path.isfile(relax_options['out:file:scorefile']):
os.remove(relax_options['out:file:scorefile'])
command = create_executable_str(PATH_TO_ROSETTA_RELAX,
args=[],
options=relax_options)
if run:
run_local_commandline(command)
# the only output we need
return relax_options['out:file:silent']
else:
return command, relax_options['out:file:silent']
def run_psiblast( sequence_filename , run = True ):
"""
Runs PSIBLAST on <sequence_filename> using the default options in
PSIBLAST_OPTIONS and returns the relevant output file: "out_ascii_pssm"
"""
root_filename = os.path.abspath( sequence_filename ).rstrip( '.fa' )
# collect the options, set the input, derive the output filenames
psiblast_options = {}
psiblast_options.update( PSIBLAST_OPTIONS )
psiblast_options['query'] = sequence_filename
for i in psiblast_options.keys():
if '__call__' in dir( psiblast_options[i] ):
psiblast_options[i] = psiblast_options[i]( root_filename )
for i in psiblast_options.keys():
if isinstance( psiblast_options[i] , str ) and os.path.isfile( psiblast_options[i] ):
psiblast_options[i] = os.path.abspath( psiblast_options[i] )
command = create_executable_str( PATH_TO_PSIBLAST , args = [] , options = psiblast_options )
if run:
run_local_commandline( command )
# the only output we need
return psiblast_options['out_ascii_pssm']
else:
# just send the command
return command , psiblast_options['out_ascii_pssm']
def run_rosetta_ddg_monomer(pdb_filename,
mut_filename,
out_filename='',
out_path='',
cleanup=True,
run=True):
root_filename = os.path.abspath(pdb_filename).rstrip('.pdb')
# hardcoded...ddg_monomer is such a painful protocol...
out_filename = ''
if '/' in root_filename:
out_filename += '/'.join(root_filename.split('/')[:-1]) + '/'
out_filename += 'ddg_predictions.out'
# clear it out if it exists, otherwise it will be appended to...
if os.path.exists(out_filename):
os.remove(out_filename)
# collect the options, set the input, derive the output filenames
ddg_monomer_options = {}
ddg_monomer_options.update(ROSETTA_DDG_MONOMER_OPTIONS)
ddg_monomer_options['in:file:s'] = pdb_filename
ddg_monomer_options['ddg::mut_file'] = mut_filename
for i in ddg_monomer_options.keys():
if '__call__' in dir(ddg_monomer_options[i]):
ddg_monomer_options[i] = ddg_monomer_options[i](root_filename)
for i in ddg_monomer_options.keys():
if isinstance(ddg_monomer_options[i], str) and os.path.isfile(
ddg_monomer_options[i]):
ddg_monomer_options[i] = os.path.abspath(ddg_monomer_options[i])
command = ''
# optionally move into the specific directory...
if out_path:
command += 'cd ' + out_path + '; ' #\n\n'
command += create_executable_str(PATH_TO_ROSETTA_DDG_MONOMER,
args=[],
options=ddg_monomer_options)
if run:
run_local_commandline(command)
# optionally cleanup
if cleanup:
print 'ddg_monomer writes useless output files, deleting these now...'
remove_intermediate_ddg_monomer_files()
# the only output we need
return out_filename
else:
return command, out_filename
def run_rosetta_rescore(silent_filename,
native_filename,
score_filename='',
run=True):
"""
Performs extraction of individual PDB structures from <silent_filename>
to <out_dir> (default to current location) using the "score" protocol
of Rosetta (built against 3.5)
Optionally specify <extra_options>
"""
root_filename = os.path.abspath(silent_filename).rstrip('.silent')
score_options = {}
score_options.update(ROSETTA_SCORE_OPTIONS)
score_options['in:file:silent'] = silent_filename
score_options[
'in:file:native'] = native_filename # required to get gdtmm scores
for i in score_options.keys():
if '__call__' in dir(score_options[i]):
score_options[i] = score_options[i](root_filename)
# necessary...
if 'out:file:scorefile' in score_options.keys(
) and not 'rescore.sc' in score_options['out:file:scorefile']:
score_options['out:file:scorefile'] = score_options[
'out:file:scorefile'].replace('.sc', '_rescore.sc')
for i in score_options.keys():
if isinstance(score_options[i], str) and os.path.isfile(
score_options[i]):
score_options[i] = os.path.abspath(score_options[i])
# ...weird Rosetta append behavior...
if os.path.isfile(score_options['out:file:scorefile']):
os.remove(score_options['out:file:scorefile'])
# default options
command = create_executable_str(PATH_TO_ROSETTA_SCORE,
args=[],
options=score_options)
if run:
run_local_commandline(command)
return score_options['out:file:scorefile']
else:
return command, score_options['out:file:scorefile']
def get_slurm_queue_status( user = SLURM_USER , header_lines = 1 , trailer_lines = 0 , only_job_status = True ):
# header_lines = 2 for FULL queue, = 5 for USER queue ("-u")
command = 'squeue'
if user:
command += ' -u ' + user
# queue_info = subprocess.Popen( command.split( ' ' ) , stdout = subprocess.PIPE , stdin = subprocess.PIPE , stderr = subprocess.STDOUT ).communicate()[0]
queue_info = run_local_commandline( command , collect_stdout = True )
# debug
# print queue_info
# simple parsing
queue_info = queue_info.split( '\n' )
if trailer_lines:
queue_info = queue_info[header_lines:-1*trailer_lines]
elif header_lines:
queue_info = queue_info[header_lines:]
queue_info = [[j for j in i.split( ' ' ) if j.strip()] for i in queue_info if i.strip()]
# debug
# print queue_info
# optionally only report the job statuses
if only_job_status:
# queue_info = [(i[0][:i[0].find( '.' )] , i[-2]) for i in queue_info]
queue_info = [(i[0] , i[4]) for i in queue_info]
# make into a dict? job ids should be unique...
queue_info = dict( queue_info )
# debug
# print queue_info
return queue_info
def run_rosetta_ddg_monomer( pdb_filename , mut_filename , out_filename = '' , out_path = '' , cleanup = True , run = True ):
root_filename = os.path.abspath( pdb_filename ).rstrip( '.pdb' )
# hardcoded...ddg_monomer is such a painful protocol...
out_filename = ''
if '/' in root_filename:
out_filename += '/'.join( root_filename.split( '/' )[:-1] ) +'/'
out_filename += 'ddg_predictions.out'
# clear it out if it exists, otherwise it will be appended to...
if os.path.exists( out_filename ):
os.remove( out_filename )
# collect the options, set the input, derive the output filenames
ddg_monomer_options = {}
ddg_monomer_options.update( ROSETTA_DDG_MONOMER_OPTIONS )
ddg_monomer_options['in:file:s'] = pdb_filename
ddg_monomer_options['ddg::mut_file'] = mut_filename
for i in ddg_monomer_options.keys():
if '__call__' in dir( ddg_monomer_options[i] ):
ddg_monomer_options[i] = ddg_monomer_options[i]( root_filename )
for i in ddg_monomer_options.keys():
if isinstance( ddg_monomer_options[i] , str ) and os.path.isfile( ddg_monomer_options[i] ):
ddg_monomer_options[i] = os.path.abspath( ddg_monomer_options[i] )
command = ''
# optionally move into the specific directory...
if out_path:
command += 'cd '+ out_path +'; '
command += create_executable_str( PATH_TO_ROSETTA_DDG_MONOMER , args = [] , options = ddg_monomer_options )
if run:
run_local_commandline( command )
# optionally cleanup
if cleanup:
print 'ddg_monomer writes useless output files, deleting these now...'
remove_intermediate_ddg_monomer_files()
# the only output we need
return out_filename
else:
return command , out_filename
def run_probe(pdb_filename, variants, probe_output_filename='', run=True):
"""
Runs PROBE on <pdb_filename> on the positions found among <variants>
using the default options in PROBE_OPTIONS and writes the output to
<probe_output_filename> (also returns this output filename)
"""
if not probe_output_filename:
probe_output_filename = os.path.abspath(pdb_filename).rstrip(
'.pdb') + '.probe_out'
# get the unique variant positions
positions = list(set([i[1:-1] for i in variants]))
positions.sort()
# generate the commands to run
# command = '#!/bin/sh\nrm ' + probe_output_filename + '\ntouch ' + probe_output_filename + '\n'
command = 'rm ' + probe_output_filename + ';touch ' + probe_output_filename + ';'
# delete any prior copy since we will append to it
for i in positions:
probe_options = {}
probe_options.update(PROBE_OPTIONS)
probe_options['out'] = pdb_filename
probe_options['Q'] = str(i)
command += create_executable_str(PATH_TO_PROBE, [],
probe_options,
probe_output_filename,
append=True) + ';' #'\n'
# run PROBE, store the output
if run:
run_local_commandline(command)
return probe_output_filename, positions
else:
# the command, well, get positions etc. too
return command, probe_output_filename, positions
def run_rosetta_rescore( silent_filename , native_filename , score_filename = '' , run = True ):
"""
Performs extraction of individual PDB structures from <silent_filename>
to <out_dir> (default to current location) using the "score" protocol
of Rosetta (built against 3.5)
Optionally specify <extra_options>
"""
root_filename = os.path.abspath( silent_filename ).rstrip( '.silent' )
score_options = {}
score_options.update( ROSETTA_SCORE_OPTIONS )
score_options['in:file:silent'] = silent_filename
score_options['in:file:native'] = native_filename # required to get gdtmm scores
for i in score_options.keys():
if '__call__' in dir( score_options[i] ):
score_options[i] = score_options[i]( root_filename )
# necessary...
if 'out:file:scorefile' in score_options.keys() and not 'rescore.sc' in score_options['out:file:scorefile']:
score_options['out:file:scorefile'] = score_options['out:file:scorefile'].replace( '.sc' , '_rescore.sc' )
for i in score_options.keys():
if isinstance( score_options[i] , str ) and os.path.isfile( score_options[i] ):
score_options[i] = os.path.abspath( score_options[i] )
# ...weird Rosetta append behavior...
if os.path.isfile( score_options['out:file:scorefile'] ):
os.remove( score_options['out:file:scorefile'] )
# default options
command = create_executable_str( PATH_TO_ROSETTA_SCORE , args = [] , options = score_options )
if run:
run_local_commandline( command )
return score_options['out:file:scorefile']
else:
return command , score_options['out:file:scorefile']
def run_slurm_job( script_filename , slurm_run_command = 'sbatch' , output_filename = 'temp_slurm.sh' ):
# optionally write a file
if not os.path.isfile( script_filename ) and isinstance( script_filename , str ):
# if os.path.isfile( output_filename ): # should really check if it exists first...
f = open( output_filename , 'w' )
f.write( script_filename )
f.close()
script_filename = output_filename
# submit it, grab the name :)
# job_id = run_and_record( slurm_run_command +' '+ script_filename , )
job_id = run_local_commandline( slurm_run_command +' '+ script_filename , collect_stdout = True )
job_id = job_id[0].split( ' ' )[-1]
return job_id
def get_pbs_queue_status( user = PBS_USER , header_lines = 5 , trailer_lines = 1 , only_job_status = True ):
# header_lines = 2 for FULL queue, = 5 for USER queue ("-u")
command = 'qstat'
if user:
command += ' -u ' + user
queue_info = run_local_commandline( command , collect_stdout = True )
# simple parsing
queue_info = queue_info.split( '\n' )[header_lines:-1*trailer_lines]
queue_info = [[j for j in i.split( ' ' ) if j.strip()] for i in queue_info]
# optionally only report the job statuses
if only_job_status:
queue_info = [(i[0][:i[0].find( '.' )] , i[-2]) if '.' in i[0] else (i[0] , i[-2]) for i in queue_info]
# make into a dict? job ids should be unique...
queue_info = dict( queue_info )
return queue_info
def get_pbs_queue_status(user=PBS_USER,
header_lines=5,
trailer_lines=1,
only_job_status=True):
# header_lines = 2 for FULL queue, = 5 for USER queue ("-u")
command = 'qstat'
if user:
command += ' -u ' + user
queue_info = run_local_commandline(command, collect_stdout=True)
# simple parsing
queue_info = queue_info.split('\n')[header_lines:-1 * trailer_lines]
queue_info = [[j for j in i.split(' ') if j.strip()] for i in queue_info]
# optionally only report the job statuses
if only_job_status:
queue_info = [(i[0][:i[0].find('.')], i[-2]) if '.' in i[0] else
(i[0], i[-2]) for i in queue_info]
# make into a dict? job ids should be unique...
queue_info = dict(queue_info)
return queue_info
def run_VIPUR_PBS(pdb_filename='',
variants_filename='',
out_path='',
write_numbering_map=True,
single_relax=True,
delete_intermediate_relax_files=True,
demo=False,
rerun_preprocessing=False):
# for the example input
if demo:
pdb_filename = PATH_TO_VIPUR + '/example_input/2C35.pdb'
variants_filename = PATH_TO_VIPUR + '/example_input/2C35.txt'
out_path = PATH_TO_VIPUR + '/example_output'
# alternatively, run on an entire directory
if not pdb_filename and not variants_filename:
# current directory
print 'no input provided, assuming you want to run on every (.pdb,.txt) file pair found in the current directory'
pdb_filename = os.getcwd()
if os.path.isdir(pdb_filename) and not variants_filename:
# assume variants_filename from pdb_filename
variants_filename = '.txt'
if os.path.isdir(pdb_filename) and variants_filename[0] == '.':
# look for file extension
# instead, run on the directory
if not out_path:
out_path = os.path.abspath(pdb_filename)
fa_filenames = [(out_path + '/') * bool(out_path) + i
for i in os.listdir(pdb_filename)
if get_file_extension(i) == 'fa']
fa_filenames = [
[i, get_root_filename(i) + variants_filename] for i in fa_filenames
if os.path.isfile(get_root_filename(i) + variants_filename)
and not os.path.isfile(get_root_filename(i) + '.pdb')
]
print 'running VIPUR on all (.pdb,' + variants_filename + ') file pairs found in ' + pdb_filename
# find .pdb files
pdb_filenames = [(out_path + '/') * bool(out_path) + i
for i in os.listdir(pdb_filename)
if get_file_extension(i) == 'pdb']
# look for pairs
pdb_filenames = [
[i, get_root_filename(i) + variants_filename]
for i in pdb_filenames
if os.path.isfile(get_root_filename(i) + variants_filename)
]
print str(len(pdb_filenames)) + ' pairs found'
print str(len(fa_filenames)) + ' pairs found for sequence only mode'
# go there...
# os.chdir( pdb_filename )
if not pdb_filenames:
if not fa_filenames:
raise IOError('!!! no (.pdb,' + variants_filename +
') file pairs found in ' + pdb_filename +
'!!?!\nAND no (.fa,' + variants_filename +
') file pairs were found...')
else:
print '...only (.fa,' + variants_filename + ') file pairs were found, running in sequence only mode'
else:
# file extension etc.
file_extension = get_file_extension(pdb_filename)
root_filename = get_root_filename(pdb_filename)
# normal execution, generalize by turning into list
pdb_filenames = []
fa_filenames = []
if file_extension == 'pdb':
pdb_filenames = [[
(out_path + '/') * bool(out_path) + pdb_filename,
(out_path + '/') * bool(out_path) + variants_filename
]]
else:
fa_filenames = [[]]
# combine all "filenames" to run into unified framework
target_proteins = [] #None]*(len( pdb_filenames ) + len( fa_filenames ))
for i in pdb_filenames:
this_out_path = get_root_filename(
i[0]) + '_VIPUR' # directory to create
target_proteins.append(i + [False, this_out_path])
for i in fa_filenames:
this_out_path = get_root_filename(
i[0]) + '_VIPUR' # directory to create
target_proteins.append(i + [True, this_out_path])
# setup environment variables BEFORE pre processing
# no need to setup a command, just run it
if PBS_ENVIRONMENT_SETUP:
print 'setting up environment variables'
run_local_commandline(PBS_ENVIRONMENT_SETUP)
# pre processing
task_summaries = []
for i in target_proteins:
# guess what the task summary filename 'would' be, if it exists, keep going...
task_summary_filename = i[3] * bool(i[3]) + '/' + get_root_filename(
i[0]).split('/')[-1] + '.task_summary'
if os.path.isfile(task_summary_filename) and not rerun_preprocessing:
print 'hmmm, ' + i[
0] + ' seems to have run preprocessing already, skipping now'
#continue # skip this one, do not add to list of tasks...?
# actually, skip running pre-processing BUT DO add it to the list of tasks
else:
task_summary_filename = run_preprocessing(
i[0],
i[1],
sequence_only=i[2],
out_path=i[3],
task_summary_filename=task_summary_filename,
write_numbering_map=write_numbering_map,
single_relax=single_relax,
pymol_environment_setup=PBS_ENVIRONMENT_SETUP)
# modify for PBS script
task_summary = load_task_summary(task_summary_filename)
for j in xrange(len(task_summary['commands'])):
command = task_summary['commands'][j]['command']
# add for relax
if task_summary['commands'][j]['feature'].replace(
'_native',
'') == 'relax' and not 'rescore' in task_summary[
'commands'][j]['feature']:
if not PBS_PARALLEL_ROSETTA_ENDING in command:
command = command.replace(ROSETTA_ENDING,
PBS_PARALLEL_ROSETTA_ENDING)
command = PBS_PARALLEL_ROSETTA_EXECUTION_COMMAND + ' ' * bool(
PBS_PARALLEL_ROSETTA_EXECUTION_COMMAND) + command
if ROSETTA_RELAX_PARALLEL_OPTIONS:
command += ' ' + ' '.join([
'-' + k + (' ' + ROSETTA_RELAX_PARALLEL_OPTIONS[k]) *
bool(ROSETTA_RELAX_PARALLEL_OPTIONS[k])
for k in ROSETTA_RELAX_PARALLEL_OPTIONS
])
# command += ' -jd2:mpi_file_buf_job_distributor false'
# command += ' -run:multiple_processes_writing_to_one_directory'
# also use the parallel options
pbs_options = 'parallel' #.update( PBS_PARALLEL_JOB_OPTIONS )
else:
pbs_options = 'serial' #.update( PBS_SERIAL_JOB_OPTIONS )
# put "cd" in front
command = ('cd ' + i[3] + ';') * bool(i[3]) + command
# modify the task summary
task_summary['commands'][j]['command'] = command
# actually write the script...
# don't worry about optional #PBS header info
# need to add the variant? no, just use the output_filename for this
script_filename = i[3] + '/' * bool(i[3]) + get_root_filename(
task_summary['commands'][j]['output_filename'].split('/')[-1]
) + '.' + task_summary['commands'][j]['feature'] + '.pbs_script.sh'
task_summary['commands'][j]['script_filename'] = script_filename
f = open(script_filename, 'w')
f.write(PBS_BASH_SCRIPT(command))
f.close()
# use the script filename as the source for any log files
# control the output and error paths
# also generate the pbs call? might as well, keep it simple
# no, uses ":" and "," characters...
task_summary['commands'][j]['queue'] = pbs_options
# rewrite the task summary
write_task_summary(task_summary, task_summary_filename)
task_summaries.append(task_summary) #_filename )
# run them all
run_VIPUR_task_summaries_PBS(
task_summaries,
single_relax=single_relax,
delete_intermediate_relax_files=delete_intermediate_relax_files)
# post processing
# this look identical!!! :)
for i in xrange(len(task_summaries)):
# always okay to rerun post processing...should not make any difference
sequence_only = target_proteins[i][2]
print '\n\n\nExtracting and Analyzing the Results:\n\n'
task_summaries[i] = run_postprocessing(task_summaries[i],
sequence_only=sequence_only)
return task_summaries
def run_VIPUR_tasks_PBS(task_summaries,
task_list,
max_pbs_tries=2,
ddg_monomer_cleanup=True,
single_relax=True,
delete_intermediate_relax_files=False):
# run the non_rescore tasks
completed = [
i for i in task_list if 'run' in task_summaries[i[0]]['commands'][i[1]]
and 'success' in task_summaries[i[0]]['commands'][i[1]]['run']
]
# should running_or_queued be saved? written to file?
running_or_queued = {}
rounds = 0
all_completed_jobs = [
] # prevents annoying bulk output, only see it the first time it completes
while not len(completed) == len(task_list):
rounds += 1
print '\n\nQUEUE MONITOR ROUND ' + str(rounds)
# debug
# print running_or_queued
# check queue status
queue_status = get_pbs_queue_status()
# update "running_or_queued" list (?)
# err, no, does not have information on which job it is...:(
#for i in queue_status.keys():
# if queue_status[i] in ['R' , 'Q']:
queue_space_occupied = len([
i for i in queue_status.values() if not i in ['C', 'R']
]) # ignore "C"ompleted jobs, "R"unning job quota are not set by us...
# if your queue system does not have a separate "R"un quota, remove 'R' from the above!
available_space = PBS_QUEUE_QUOTA - queue_space_occupied
# launch next jobs in available slots
if available_space:
print str(queue_space_occupied
) + ' jobs queued or running, could submit up to ' + str(
available_space) + ' more'
# choose the next job
jobs_to_run = [
i for i in task_list
if not i in completed and not i in running_or_queued.values()
and not ('run' in task_summaries[i[0]]['commands'][i[1]] and
('success' in task_summaries[i[0]]['commands'][
i[1]]['run'] or 'failure' in task_summaries[i[0]]
['commands'][i[1]]['run']))
]
print str(
len(jobs_to_run)
) + ' jobs still need to finish (after the currently running jobs complete)'
# only the next few
for i in jobs_to_run[:available_space]:
command_dict = task_summaries[i[0]]['commands'][i[1]]
# write scripts as part of pre processing?...yeah...
# write the command to a script
#script_filename = command_dict['out_path'] +'/'*bool( command_dict['out_path'] )+
#script_filename = command_dict['script_filename']
# if its a rescore and relax jobs were separated, need to recombine them!
if 'rescore' in command_dict['feature']:
# combine the individual relax runs
#relax_commands = [i for i in task_summary['commands'] if i['feature'].replace( '_native' , '' ) == 'relax']
#silent_filenames = [j['output_filename'] for j in relax_commands if j['variant'] == i['variant'] and 'run' in j.keys() and j['run'] == 'success']
silent_filenames = [
j['output_filename']
for j in task_summaries[i[0]]['commands']
if j['feature'].replace('_native', '') == 'relax'
and j['variant'] == command_dict['variant']
and 'run' in j.keys() and 'success' in j['run']
]
# actually need to identify the combined_silent_filename, be sure the relax files have not already been merged
# which variant
target_variant = [
j for j in task_summaries[i[0]]['variants'].keys()
if j.split('_')[-1] == command_dict['variant']
and j.split('_')[0] in command_dict['command']
]
if not target_variant:
# its native
combined_silent_filename = task_summaries[
i[0]]['other']['combined_native_silent_filename']
combined_score_filename = task_summaries[
i[0]]['other']['combined_native_score_filename']
elif len(target_variant) > 1:
raise Exception(
'??? found more than on matching variant ???\n' +
', '.join(target_variant))
else:
# found it
combined_silent_filename = task_summaries[
i[0]]['variants'][
target_variant[0]]['combined_silent_filename']
combined_score_filename = task_summaries[
i[0]]['variants'][
target_variant[0]]['combined_score_filename']
#if not single_relax: # AND post processing has not already be run...scan for the combined silent file
if not single_relax and not os.path.isfile(
combined_silent_filename):
if not len(silent_filenames
) == ROSETTA_RELAX_OPTIONS['nstruct']:
raise Exception(
'??? somehow the matching relax run(s) has failed ???\n'
+ str(i))
score_filenames = [
j.replace('.silent', '.sc')
for j in silent_filenames
]
merge_rosetta_relax_output(
silent_filenames,
combined_silent_filename,
score_filenames,
combined_score_filename,
delete_old_files=delete_intermediate_relax_files)
# rescore already knows the proper filename
else:
# just a single match for each
# output filename should be correct as is :)
None
# submit this script using a queue command
# generate it here instead
pbs_options = {}
if command_dict['queue'] == 'parallel':
pbs_options.update(PBS_PARALLEL_JOB_OPTIONS)
elif command_dict['queue'] == 'serial':
pbs_options.update(PBS_SERIAL_JOB_OPTIONS)
# make sure they are satisfied
script_filename = command_dict['script_filename']
for k in pbs_options.keys():
if '__call__' in dir(pbs_options[k]):
pbs_options[k] = pbs_options[k](script_filename)
pbs_command = create_executable_str('qsub', [script_filename],
pbs_options)
new_job_id = run_local_commandline(pbs_command,
collect_stdout=True)
new_job_id = new_job_id.strip()
if '.' in new_job_id:
new_job_id = new_job_id[:new_job_id.find('.')]
print 'submitted ' + new_job_id
# save the job id
# assume its queue
running_or_queued[new_job_id] = i
else:
print 'no new \"positions\" are available'
# debug, need to know
running_jobs = len([i for i in queue_status.values() if i in ['R']])
if running_jobs:
print str(
running_jobs
) + ' are still running...(excluding the jobs just submitted and including your other jobs)'
# assess outcome of completed jobs
for job_id in sorted(
queue_status.keys()): # sort in numerical order, right?
# debug
if not job_id in all_completed_jobs:
print '\t' + job_id, queue_status[
job_id] # , job_id in running_or_queued.keys()
# could just skip it all now?
if queue_status[
job_id] == 'C' and job_id in running_or_queued.keys():
task_id = running_or_queued[job_id][0]
command_index = running_or_queued[job_id][1]
command_dict = task_summaries[task_id]['commands'][
command_index]
check_successful = determine_check_successful_function(
command_dict, single_relax=single_relax)
success = check_successful(command_dict)
failure_summary = ''
if isinstance(success, bool):
complete = success
elif len(success) > 1 and isinstance(success[0], bool):
complete = success[0]
failure_summary += ' ' + ';'.join(
[str(j) for j in success[1:]]) + ' '
print complete, failure_summary, 'try again?' * bool(
not complete) # debug
# track the number of attempts?
# try until failure - how many times?
tries = 0
if 'run' in command_dict.keys(
) and command_dict['run'] and not 'success' in command_dict[
'run'] and not 'failure' in command_dict['run']:
tries = int(command_dict['run'])
tries += 1
print tries, 'attempts so far' # debug
if tries >= max_pbs_tries:
# its a failure
print job_id + ' completed successfully' * complete + (
' failed with ' + str(tries) +
' attempts') * (not complete)
failure_summary = 'success' * complete + (
str(tries) + ' tries;failure ' +
failure_summary) * (not complete)
elif complete:
print job_id + ' completed successfully'
failure_summary = 'success' #+ str( tries ) + ' tries'
else:
# record the number of tries
print job_id + ' completed' + ' successfully' * complete
failure_summary = str(tries)
# update the record
print 'updating with: ' + failure_summary # debug
task_summaries[task_id]['commands'][command_index][
'run'] = failure_summary
# optionally cleanup
if ddg_monomer_cleanup and command_dict[
'feature'] == 'ddg_monomer': #'ddg' in i['output_filename']:
print 'ddg_monomer writes useless output files, deleting these now...'
remove_intermediate_ddg_monomer_files()
# jobs that have since been completed - consider them complete?
completed.append(
running_or_queued[job_id]) # good, so this grows
del running_or_queued[job_id]
# remove jobs to run?
# print 'updating the status...' # debug
# write out "completed"? or "running_or_queued"?
if queue_status[job_id] == 'C' and not job_id in all_completed_jobs:
all_completed_jobs.append(
job_id) # prevent redundant update info
# update task_summaries e.g. write them!
# modified: so the task summary records its own name...bah!
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError(
'should input the task summary filename (not the summary itself)...'
)
else:
# write it out
print 'updating: ' + i['filenames']['task_summary_filename']
write_task_summary(i, i['filenames']['task_summary_filename'])
# pause...
print '\n', len(completed), 'completed', len(
task_list), 'tasks remaining' # debug
if len(completed) <= len(task_list): # no need for edge-case end wait
print 'waiting ' + str(PBS_QUEUE_MONITOR_DELAY) + 's...'
time.sleep(PBS_QUEUE_MONITOR_DELAY)
# return anything?
# write one last time?
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError(
'should input the task summary filename (not the summary itself)...'
)
else:
# write it out
write_task_summary(i, i['filenames']['task_summary_filename'])
def run_rosetta_relax(pdb_filename,
extra_options={},
run=True,
parallel=ROSETTA_RELAX_PARALLEL):
root_filename = pdb_filename.rstrip('.pdb')
# collect the options, set the input, derive the output filenames
relax_options = {}
relax_options.update(ROSETTA_RELAX_OPTIONS)
relax_options.update(extra_options)
relax_options['s'] = pdb_filename
relax_options['native'] = pdb_filename # required to get gdtmm scores
for i in relax_options.keys():
if '__call__' in dir(relax_options[i]):
relax_options[i] = relax_options[i](root_filename)
# ...weird Rosetta append behavior...
# if os.path.isfile( relax_options['out:file:silent'] ):
# os.remove( relax_options['out:file:silent'] )
# if os.path.isfile( relax_options['out:file:scorefile'] ):
# os.remove( relax_options['out:file:scorefile'] )
# for njc parallelization
nstruct = int(relax_options.get('nstruct', '0'))
parallel = int(parallel)
tmp_file = None
if nstruct > 1 and parallel > 1:
relax_options['nstruct'] = 1 #TODO: Add chunking option?
score_filename = relax_options['out:file:scorefile']
silent_filename = relax_options['out:file:silent']
if 'run:jran' in relax_options:
restoreJran = True
jran = int(relax_options['run:jran'])
else:
restoreJran = False
jran = 123
tmp_file = tempfile.NamedTemporaryFile(delete=False)
print 'Parallel relax commands are in ' + tmp_file.name
for s in xrange(nstruct):
tag = '_%05d' % s
relax_options['run:jran'] = jran * nstruct + s
relax_options['out:file:scorefile'] = score_filename + tag
relax_options['out:file:silent'] = silent_filename + tag
print >> tmp_file, create_executable_str(
PATH_TO_ROSETTA_RELAX, args=[], options=relax_options
) + " > %s 2>&1; echo '[[VIPURLOG]]' %s %d" % (
(silent_filename + tag).replace('silent_',
'log_'), pdb_filename, s + 1)
tmp_file.close()
# the "find ... | xargs ..." idiom is used just in case nstruct is ever a *very* large number.
command = '''\
parallel -j %d -a %s
find . -name '%s_[0-9]*[0-9]' | xargs cat | awk 'NR == 1 || $2 != "score" {print $0}' > %s
find . -name '%s_[0-9]*[0-9]' | xargs rm
find . -name '%s_[0-9]*[0-9]' | xargs cat | awk 'NR <= 2 || !($2 == "score" || $1 == "SEQUENCE:") {print $0}' > %s
find . -name '%s_[0-9]*[0-9]' | xargs rm
''' % (parallel, tmp_file.name, score_filename, score_filename, score_filename,
silent_filename, silent_filename, silent_filename)
print 'Parallel relax driver command:', command
# restore option values
relax_options['nstruct'] = str(nstruct)
relax_options['out:file:scorefile'] = score_filename
relax_options['out:file:silent'] = silent_filename
if restoreJran:
relax_options['run:jran'] = jran
if run:
return (command, tmp_file.name, score_filename, silent_filename)
if tmp_file:
os.unlink(tmp_file.name)
else:
command = create_executable_str(PATH_TO_ROSETTA_RELAX,
args=[],
options=relax_options)
if run:
run_local_commandline(command)
# the only output we need
# return relax_options['out:file:scorefile']
return relax_options['out:file:silent']
def run_VIPUR_tasks_PBS( task_summaries , task_list , max_pbs_tries = 2 , ddg_monomer_cleanup = True , single_relax = True , delete_intermediate_relax_files = False ):
# run the non_rescore tasks
completed = [i for i in task_list if 'run' in task_summaries[i[0]]['commands'][i[1]] and 'success' in task_summaries[i[0]]['commands'][i[1]]['run']]
# should running_or_queued be saved? written to file?
running_or_queued = {}
rounds = 0
all_completed_jobs = [] # prevents annoying bulk output, only see it the first time it completes
while not len( completed ) == len( task_list ):
rounds += 1
print '\n\nQUEUE MONITOR ROUND ' + str( rounds )
# debug
# print running_or_queued
# check queue status
queue_status = get_pbs_queue_status()
# update "running_or_queued" list (?)
# err, no, does not have information on which job it is...:(
#for i in queue_status.keys():
# if queue_status[i] in ['R' , 'Q']:
queue_space_occupied = len( [i for i in queue_status.values() if not i in ['C' , 'R']] ) # ignore "C"ompleted jobs, "R"unning job quota are not set by us...
# if your queue system does not have a separate "R"un quota, remove 'R' from the above!
available_space = PBS_QUEUE_QUOTA - queue_space_occupied
# launch next jobs in available slots
if available_space:
print str( queue_space_occupied ) + ' jobs queued or running, could submit up to ' + str( available_space ) + ' more'
# choose the next job
jobs_to_run = [i for i in task_list if
not i in completed and
not i in running_or_queued.values() and
not ( 'run' in task_summaries[i[0]]['commands'][i[1]] and
('success' in task_summaries[i[0]]['commands'][i[1]]['run'] or
'failure' in task_summaries[i[0]]['commands'][i[1]]['run']) )
]
print str( len( jobs_to_run ) ) + ' jobs still need to finish (after the currently running jobs complete)'
# only the next few
for i in jobs_to_run[:available_space]:
command_dict = task_summaries[i[0]]['commands'][i[1]]
# write scripts as part of pre processing?...yeah...
# write the command to a script
#script_filename = command_dict['out_path'] +'/'*bool( command_dict['out_path'] )+
#script_filename = command_dict['script_filename']
# if its a rescore and relax jobs were separated, need to recombine them!
if 'rescore' in command_dict['feature']:
# combine the individual relax runs
#relax_commands = [i for i in task_summary['commands'] if i['feature'].replace( '_native' , '' ) == 'relax']
#silent_filenames = [j['output_filename'] for j in relax_commands if j['variant'] == i['variant'] and 'run' in j.keys() and j['run'] == 'success']
silent_filenames = [j['output_filename'] for j in task_summaries[i[0]]['commands'] if
j['feature'].replace( '_native' , '' ) == 'relax' and
j['variant'] == command_dict['variant'] and
'run' in j.keys() and
'success' in j['run']
]
# actually need to identify the combined_silent_filename, be sure the relax files have not already been merged
# which variant
target_variant = [j for j in task_summaries[i[0]]['variants'].keys() if j.split( '_' )[-1] == command_dict['variant'] and j.split( '_' )[0] in command_dict['command']]
if not target_variant:
# its native
combined_silent_filename = task_summaries[i[0]]['other']['combined_native_silent_filename']
combined_score_filename = task_summaries[i[0]]['other']['combined_native_score_filename']
elif len( target_variant ) > 1:
raise Exception( '??? found more than on matching variant ???\n' + ', '.join( target_variant ) )
else:
# found it
combined_silent_filename = task_summaries[i[0]]['variants'][target_variant[0]]['combined_silent_filename']
combined_score_filename = task_summaries[i[0]]['variants'][target_variant[0]]['combined_score_filename']
#if not single_relax: # AND post processing has not already be run...scan for the combined silent file
if not single_relax and not os.path.isfile( combined_silent_filename ):
if not len( silent_filenames ) == ROSETTA_RELAX_OPTIONS['nstruct']:
raise Exception( '??? somehow the matching relax run(s) has failed ???\n' + str( i ) )
score_filenames = [j.replace( '.silent' , '.sc' ) for j in silent_filenames]
merge_rosetta_relax_output( silent_filenames , combined_silent_filename , score_filenames , combined_score_filename , delete_old_files = delete_intermediate_relax_files )
# rescore already knows the proper filename
else:
# just a single match for each
# output filename should be correct as is :)
None
# submit this script using a queue command
# generate it here instead
pbs_options = {}
if command_dict['queue'] == 'parallel':
pbs_options.update( PBS_PARALLEL_JOB_OPTIONS )
elif command_dict['queue'] == 'serial':
pbs_options.update( PBS_SERIAL_JOB_OPTIONS )
# make sure they are satisfied
script_filename = command_dict['script_filename']
for k in pbs_options.keys():
if '__call__' in dir( pbs_options[k] ):
pbs_options[k] = pbs_options[k]( script_filename )
pbs_command = create_executable_str( 'qsub' , [script_filename] , pbs_options )
new_job_id = run_local_commandline( pbs_command , collect_stdout = True )
new_job_id = new_job_id.strip()
if '.' in new_job_id:
new_job_id = new_job_id[:new_job_id.find( '.' )]
print 'submitted ' + new_job_id
# save the job id
# assume its queue
running_or_queued[new_job_id] = i
else:
print 'no new \"positions\" are available'
# debug, need to know
running_jobs = len( [i for i in queue_status.values() if i in ['R']] )
if running_jobs:
print str( running_jobs ) + ' are still running...(excluding the jobs just submitted and including your other jobs)'
# assess outcome of completed jobs
for job_id in sorted( queue_status.keys() ): # sort in numerical order, right?
# debug
if not job_id in all_completed_jobs:
print '\t'+ job_id , queue_status[job_id]# , job_id in running_or_queued.keys()
# could just skip it all now?
if queue_status[job_id] == 'C' and job_id in running_or_queued.keys():
task_id = running_or_queued[job_id][0]
command_index = running_or_queued[job_id][1]
command_dict = task_summaries[task_id]['commands'][command_index]
check_successful = determine_check_successful_function( command_dict , single_relax = single_relax )
success = check_successful( command_dict )
failure_summary = ''
if isinstance( success , bool ):
complete = success
elif len( success ) > 1 and isinstance( success[0] , bool ):
complete = success[0]
failure_summary += ' '+ ';'.join( [str( j ) for j in success[1:]] ) +' '
print complete , failure_summary , 'try again?'*bool( not complete ) # debug
# track the number of attempts?
# try until failure - how many times?
tries = 0
if 'run' in command_dict.keys() and command_dict['run'] and not 'success' in command_dict['run'] and not 'failure' in command_dict['run']:
tries = int( command_dict['run'] )
tries += 1
print tries , 'attempts so far' # debug
if tries >= max_pbs_tries:
# its a failure
print job_id + ' completed successfully'*complete + (' failed with ' + str( tries ) + ' attempts')*(not complete)
failure_summary = 'success'*complete + (str( tries ) +' tries;failure ' + failure_summary)*(not complete)
elif complete:
print job_id + ' completed successfully'
failure_summary = 'success' #+ str( tries ) + ' tries'
else:
# record the number of tries
print job_id + ' completed' + ' successfully'*complete
failure_summary = str( tries )
# update the record
print 'updating with: ' + failure_summary # debug
task_summaries[task_id]['commands'][command_index]['run'] = failure_summary
# optionally cleanup
if ddg_monomer_cleanup and command_dict['feature'] == 'ddg_monomer':#'ddg' in i['output_filename']:
print 'ddg_monomer writes useless output files, deleting these now...'
remove_intermediate_ddg_monomer_files()
# jobs that have since been completed - consider them complete?
completed.append( running_or_queued[job_id] ) # good, so this grows
del running_or_queued[job_id]
# remove jobs to run?
# print 'updating the status...' # debug
# write out "completed"? or "running_or_queued"?
if queue_status[job_id] == 'C' and not job_id in all_completed_jobs:
all_completed_jobs.append( job_id ) # prevent redundant update info
# update task_summaries e.g. write them!
# modified: so the task summary records its own name...bah!
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
print 'updating: ' + i['filenames']['task_summary_filename']
write_task_summary( i , i['filenames']['task_summary_filename'] )
# pause...
print '\n' , len( completed ) , 'completed' , len( task_list ) , 'tasks remaining' # debug
if len( completed ) <= len( task_list ): # no need for edge-case end wait
print 'waiting ' + str( PBS_QUEUE_MONITOR_DELAY ) +'s...'
time.sleep( PBS_QUEUE_MONITOR_DELAY )
# return anything?
# write one last time?
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
write_task_summary( i , i['filenames']['task_summary_filename'] )
def create_variant_protein_structures( pdb_filename , variants , chain , use_pyrosetta = USE_PYROSETTA ):
# make sure the variants have been filtered
if use_pyrosetta:
# load the PDB as a pose
pose = pose_from_pdb( pdb_filename )
failed = {}
root_filename = pdb_filename.rstrip( 'pdb' )
# currently cannot handle multi-chain input
# handle this before VIPUR
if pose.chain( pose.total_residue() ) > 1:
print 'CANNOT currently handle multi-chain PDBs (as pose), using PyMOL instead!'
if PATH_TO_PYMOL:
create_variant_protein_structures( pdb_filename , variants , chain , use_pyrosetta = False )
return
else:
faulty = 'clean before VIPUR, cannot handle multi-chain PDBs'
failed[faulty] = variants
elif not pose.pdb_info().chain( 1 ) == chain:
print '...not sure what it happening, you wanted chain ' + chain + ' but VIPUR found chain ' + pose.chain( 1 ) + ', skipping this entire sample!'
faulty = 'clean before VIPUR, improper chain ID'
failed[faulty] = variants
# in case this condition is found:
faulty = 'could not load position from PDB'
variant_structures = []
for variation in variants:
# make a copy
test_pose = Pose()
test_pose.assign( pose )
native = variation[0]
position = variation[1:-1]
mutant = variation[-1]
# make sure the position was loaded
icode = ' ' # default...this could cause problems...
if not position[-1].isdigit():
icode = position[-1]
position = position[:-1]
position = int( position )
if not test_pose.pdb_info().pdb2pose( chain , position , icode ):
if faulty in failed.keys():
failed[faulty].append( variation )
else:
failed[faulty] = [variation]
break # stop the loop
position = test_pose.pdb_info().pdb2pose( chain , position , icode )
# simple, use a mover to make the change
# appears to have trouble with N terminal variants since it uses "replace_residue"
# code is available that does not have this problem, however reloading into Rosetta with accurately determine the position of these atoms
make_variant = MutateResidue( position , mutant )
make_variant.apply( test_pose )
# write out
out_filename = self.root_filename +'.chain_'+ chain +'_'+ variation +'.pdb'
variant_structures.append( out_filename )
test_pose.dump_pdb( out_filename )
print 'generated ' + variation + ' variant structure and wrote to ' + out_filename
return variant_structures
else:
# use the pymol script
#for variants in self.variants['permissible']:
# use default output naming
# create command explicitly here, slightly different
root_filename = pdb_filename.rstrip( '.pdb' )
command = PATH_TO_PYMOL + ' -qcr ' + PATH_TO_VIPUR + '/pymol_make_variant_structure.py -- -p ' + pdb_filename + ' -m ' + ','.join( variants ) + ' -c ' + chain + ' -r ' + root_filename
# print command
run_local_commandline( command )
# reconstruct the names
variant_structures = [root_filename + '.chain_' + chain +'_'+ i +'.pdb' for i in variants]
# verify they have been made
if [None for i in variant_structures if not os.path.isfile( i )]:
raise IOError( 'could not make variant protein structures,\ntry checking the input PDB file or the pymol script pymol_make_variant_structure.py' )
return variant_structures
def run_rosetta_relax( pdb_filename , extra_options = {} , run = True , parallel = ROSETTA_RELAX_PARALLEL ):
root_filename = pdb_filename.rstrip( '.pdb' )
# collect the options, set the input, derive the output filenames
relax_options = {}
relax_options.update( ROSETTA_RELAX_OPTIONS )
relax_options.update( extra_options )
relax_options['s'] = pdb_filename
relax_options['native'] = pdb_filename # required to get gdtmm scores
for i in relax_options.keys():
if '__call__' in dir( relax_options[i] ):
relax_options[i] = relax_options[i]( root_filename )
# ...weird Rosetta append behavior...
# if os.path.isfile( relax_options['out:file:silent'] ):
# os.remove( relax_options['out:file:silent'] )
# if os.path.isfile( relax_options['out:file:scorefile'] ):
# os.remove( relax_options['out:file:scorefile'] )
# for njc parallelization
nstruct = int( relax_options.get( 'nstruct' , '0' ) )
parallel = int( parallel )
tmp_file = None
if nstruct > 1 and parallel > 1:
relax_options['nstruct'] = 1 #TODO: Add chunking option?
score_filename = relax_options['out:file:scorefile']
silent_filename = relax_options['out:file:silent']
if 'run:jran' in relax_options:
restoreJran = True
jran = int( relax_options['run:jran'] )
else:
restoreJran = False
jran = 123
tmp_file = tempfile.NamedTemporaryFile( delete = False )
print 'Parallel relax commands are in ' + tmp_file.name
for s in xrange( nstruct ):
tag = '_%05d' % s
relax_options['run:jran'] = jran*nstruct + s
relax_options['out:file:scorefile'] = score_filename + tag
relax_options['out:file:silent'] = silent_filename + tag
print >>tmp_file , create_executable_str( PATH_TO_ROSETTA_RELAX , args = [] , options = relax_options ) + " > %s 2>&1; echo '[[VIPURLOG]]' %s %d" % ((silent_filename + tag).replace( 'silent_' , 'log_' ) , pdb_filename , s + 1 )
tmp_file.close()
# the "find ... | xargs ..." idiom is used just in case nstruct is ever a *very* large number.
command = '''\
parallel -j %d -a %s
find . -name '%s_[0-9]*[0-9]' | xargs cat | awk 'NR == 1 || $2 != "score" {print $0}' > %s
find . -name '%s_[0-9]*[0-9]' | xargs rm
find . -name '%s_[0-9]*[0-9]' | xargs cat | awk 'NR <= 2 || !($2 == "score" || $1 == "SEQUENCE:") {print $0}' > %s
find . -name '%s_[0-9]*[0-9]' | xargs rm
''' % (parallel , tmp_file.name , score_filename , score_filename , score_filename , silent_filename , silent_filename , silent_filename)
print 'Parallel relax driver command:', command
# restore option values
relax_options['nstruct'] = str( nstruct )
relax_options['out:file:scorefile'] = score_filename
relax_options['out:file:silent'] = silent_filename
if restoreJran:
relax_options['run:jran'] = jran
if run:
return (command , tmp_file.name , score_filename , silent_filename)
if tmp_file:
os.unlink( tmp_file.name )
else:
command = create_executable_str( PATH_TO_ROSETTA_RELAX , args = [] , options = relax_options )
if run:
run_local_commandline( command )
# command = create_executable_str( PATH_TO_ROSETTA_RELAX , args = [] , options = relax_options )
# run_local_commandline( command )
# the only output we need
# return relax_options['out:file:scorefile']
return relax_options['out:file:silent']
def run_VIPUR_PBS( pdb_filename = '' , variants_filename = '' ,
out_path = '' , write_numbering_map = True ,
single_relax = True , delete_intermediate_relax_files = True ,
demo = False , rerun_preprocessing = False ):
# for the example input
if demo:
pdb_filename = PATH_TO_VIPUR + '/example_input/2C35.pdb'
variants_filename = PATH_TO_VIPUR + '/example_input/2C35.txt'
out_path = PATH_TO_VIPUR + '/example_output'
# alternatively, run on an entire directory
if not pdb_filename and not variants_filename:
# current directory
print 'no input provided, assuming you want to run on every (.pdb,.txt) file pair found in the current directory'
pdb_filename = os.getcwd()
if os.path.isdir( pdb_filename ) and not variants_filename:
# assume variants_filename from pdb_filename
variants_filename = '.txt'
if os.path.isdir( pdb_filename ) and variants_filename[0] == '.':
# look for file extension
# instead, run on the directory
if not out_path:
out_path = os.path.abspath( pdb_filename )
fa_filenames = [(out_path +'/')*bool( out_path ) + i for i in os.listdir( pdb_filename ) if get_file_extension( i ) == 'fa']
fa_filenames = [[i , get_root_filename( i ) + variants_filename] for i in fa_filenames if os.path.isfile( get_root_filename( i ) + variants_filename ) and not os.path.isfile( get_root_filename( i ) + '.pdb' )]
print 'running VIPUR on all (.pdb,' + variants_filename + ') file pairs found in ' + pdb_filename
# find .pdb files
pdb_filenames = [(out_path +'/')*bool( out_path ) + i for i in os.listdir( pdb_filename ) if get_file_extension( i ) == 'pdb']
# look for pairs
pdb_filenames = [[i , get_root_filename( i ) + variants_filename] for i in pdb_filenames if os.path.isfile( get_root_filename( i ) + variants_filename )]
print str( len( pdb_filenames ) ) + ' pairs found'
print str( len( fa_filenames ) ) + ' pairs found for sequence only mode'
# go there...
# os.chdir( pdb_filename )
if not pdb_filenames:
if not fa_filenames:
raise IOError( '!!! no (.pdb,' + variants_filename + ') file pairs found in ' + pdb_filename + '!!?!\nAND no (.fa,' + variants_filename + ') file pairs were found...' )
else:
print '...only (.fa,' + variants_filename + ') file pairs were found, running in sequence only mode'
else:
# file extension etc.
file_extension = get_file_extension( pdb_filename )
root_filename = get_root_filename( pdb_filename )
# normal execution, generalize by turning into list
pdb_filenames = []
fa_filenames = []
if file_extension == 'pdb':
pdb_filenames = [[(out_path +'/')*bool( out_path ) + pdb_filename , (out_path +'/')*bool( out_path ) + variants_filename]]
else:
fa_filenames = [[]]
# combine all "filenames" to run into unified framework
target_proteins = []#None]*(len( pdb_filenames ) + len( fa_filenames ))
for i in pdb_filenames:
this_out_path = get_root_filename( i[0] ) +'_VIPUR' # directory to create
target_proteins.append( i + [False , this_out_path] )
for i in fa_filenames:
this_out_path = get_root_filename( i[0] ) +'_VIPUR' # directory to create
target_proteins.append( i + [True , this_out_path] )
# setup environment variables BEFORE pre processing
# no need to setup a command, just run it
if PBS_ENVIRONMENT_SETUP:
print 'setting up environment variables'
run_local_commandline( PBS_ENVIRONMENT_SETUP )
# pre processing
task_summaries = []
for i in target_proteins:
# guess what the task summary filename 'would' be, if it exists, keep going...
task_summary_filename = i[3]*bool( i[3] ) +'/'+ get_root_filename( i[0] ).split( '/' )[-1] + '.task_summary'
if os.path.isfile( task_summary_filename ) and not rerun_preprocessing:
print 'hmmm, ' + i[0] + ' seems to have run preprocessing already, skipping now'
#continue # skip this one, do not add to list of tasks...?
# actually, skip running pre-processing BUT DO add it to the list of tasks
else:
task_summary_filename = run_preprocessing( i[0] , i[1] ,
sequence_only = i[2] , out_path = i[3] ,
task_summary_filename = task_summary_filename ,
write_numbering_map = write_numbering_map , single_relax = single_relax ,
pymol_environment_setup = PBS_ENVIRONMENT_SETUP )
# modify for PBS script
task_summary = load_task_summary( task_summary_filename )
for j in xrange( len( task_summary['commands'] ) ):
command = task_summary['commands'][j]['command']
# add for relax
if task_summary['commands'][j]['feature'].replace( '_native' , '' ) == 'relax' and not 'rescore' in task_summary['commands'][j]['feature']:
if not PBS_PARALLEL_ROSETTA_ENDING in command:
command = command.replace( ROSETTA_ENDING , PBS_PARALLEL_ROSETTA_ENDING )
command = PBS_PARALLEL_ROSETTA_EXECUTION_COMMAND + ' '*bool( PBS_PARALLEL_ROSETTA_EXECUTION_COMMAND ) + command
if ROSETTA_RELAX_PARALLEL_OPTIONS:
command += ' '+ ' '.join( ['-'+ k + (' '+ ROSETTA_RELAX_PARALLEL_OPTIONS[k])*bool( ROSETTA_RELAX_PARALLEL_OPTIONS[k] ) for k in ROSETTA_RELAX_PARALLEL_OPTIONS] )
# command += ' -jd2:mpi_file_buf_job_distributor false'
# command += ' -run:multiple_processes_writing_to_one_directory'
# also use the parallel options
pbs_options = 'parallel'#.update( PBS_PARALLEL_JOB_OPTIONS )
else:
pbs_options = 'serial'#.update( PBS_SERIAL_JOB_OPTIONS )
# put "cd" in front
command = ('cd '+ i[3] +';')*bool( i[3] ) + command
# modify the task summary
task_summary['commands'][j]['command'] = command
# actually write the script...
# don't worry about optional #PBS header info
# need to add the variant? no, just use the output_filename for this
script_filename = i[3] + '/'*bool( i[3] ) + get_root_filename( task_summary['commands'][j]['output_filename'].split( '/' )[-1] ) +'.'+ task_summary['commands'][j]['feature'] + '.pbs_script.sh'
task_summary['commands'][j]['script_filename'] = script_filename
f = open( script_filename , 'w' )
f.write( PBS_BASH_SCRIPT( command ) )
f.close()
# use the script filename as the source for any log files
# control the output and error paths
# also generate the pbs call? might as well, keep it simple
# no, uses ":" and "," characters...
task_summary['commands'][j]['queue'] = pbs_options
# rewrite the task summary
write_task_summary( task_summary , task_summary_filename )
task_summaries.append( task_summary )#_filename )
# run them all
run_VIPUR_task_summaries_PBS( task_summaries , single_relax = single_relax , delete_intermediate_relax_files = delete_intermediate_relax_files )
# post processing
# this look identical!!! :)
for i in xrange( len( task_summaries ) ):
# always okay to rerun post processing...should not make any difference
sequence_only = target_proteins[i][2]
print '\n\n\nExtracting and Analyzing the Results:\n\n'
task_summaries[i] = run_postprocessing( task_summaries[i] , sequence_only = sequence_only )
return task_summaries
def run_VIPUR_tasks_PBS( task_summaries , task_list , max_pbs_tries = 2 , ddg_monomer_cleanup = True , single_relax = False , delete_intermediate_relax_files = False ):
# run the non_rescore tasks
completed = [i for i in task_list if 'run' in task_summaries[i[0]]['commands'][i[1]] and 'success' in task_summaries[i[0]]['commands'][i[1]]['run']]
# should running_or_queued be saved? written to file?
running_or_queued = {}
while not len( completed ) == len( task_list ):
# check queue status
queue_status = get_pbs_queue_status()
# update "running_or_queued" list (?)
# err, no, does not have information on which job it is...:(
#for i in queue_status.keys():
# if queue_status[i] in ['R' , 'Q']:
queue_space_occupied = len( [i for i in queue_status.values() if not i in ['C' , 'R']] ) # ignore "C"ompleted jobs, "R"unning job quota are not set by us...
# if your queue system does not have a separate "R"un quota, remove 'R' from the above!
available_space = queue_space_occupied - PBS_QUEUE_QUOTA
# launch next jobs in available slots
if available_space:
# choose the next job
jobs_to_run = [i for i in task_list if
not i in completed and
not i in running_or_queued.values() and
not ( 'run' in task_summaries[i[0]]['commands'][i[1]] and
('success' in task_summaries[i[0]]['commands'][i[1]]['run'] or
'failure' in task_summaries[i[0]]['commands'][i[1]]['run']) )
]
for i in jobs_to_run:
command_dict = task_summaries[i[0]]['commands'][i[1]]
# write scripts as part of pre processing?...yeah...
# write the command to a script
#script_filename = command_dict['out_path'] +'/'*bool( command_dict['out_path'] )+
#script_filename = command_dict['script_filename']
# if its a rescore and relax jobs were separated, need to recombine them!
if 'rescore' in command_dict['feature']:
# combine the individual relax runs
#relax_commands = [i for i in task_summary['commands'] if i['feature'].replace( '_native' , '' ) == 'relax']
#silent_filenames = [j['output_filename'] for j in relax_commands if j['variant'] == i['variant'] and 'run' in j.keys() and j['run'] == 'success']
silent_filenames = [j['output_filename'] for j in task_summaries[i[0]]['commands'] if
j['feature'].replace( '_native' , '' ) == 'relax' and
j['variant'] == command_dict['variant'] and
'run' in j.keys() and
'success' in j['run']
]
# actually need to identify the combined_silent_filename, be sure the relax files have not already been merged
# which variant
target_variant = [j for j in task_summaries[i[0]]['variants'].keys() if j.split( '_' )[-1] == command_dict['variant'] and j.split( '_' )[0] in command_dict['command']]
if not target_variant:
# its native
combined_silent_filename = task_summaries[i[0]]['other']['combined_native_silent_filename']
combined_score_filename = task_summaries[i[0]]['other']['combined_native_score_filename']
elif len( target_variant ) > 1:
raise Exception( '??? found more than on matching variant ???\n' + ', '.join( target_variant ) )
else:
# found it
combined_silent_filename = task_summaries[i[0]]['variants'][target_variant[0]]['combined_silent_filename']
combined_score_filename = task_summaries[i[0]]['variants'][target_variant[0]]['combined_score_filename']
#if not single_relax: # AND post processing has not already be run...scan for the combined silent file
if not single_relax and not os.path.isfile( combined_silent_filename ):
if not len( silent_filenames ) == ROSETTA_RELAX_OPTIONS['nstruct']:
raise Exception( '??? somehow the matching relax run(s) has failed ???\n' + str( i ) )
score_filenames = [j.replace( '.silent' , '.sc' ) for j in silent_filenames]
merge_rosetta_relax_output( silent_filenames , combined_silent_filename , score_filenames , combined_score_filename , delete_old_files = delete_intermediate_relax_files )
# rescore already knows the proper filename
else:
# just a single match for each
# output filename should be correct as is :)
None
# submit this script using a queue command
pbs_command = command_dict['qsub_command'] # SHOULD already have an abspath to the script
new_job_id = run_local_commandline( pbs_command , collect_stdout = True )
new_job_id = new_job_id[:new_job_id.find( '.' )]
# save the job id
# assume its queue
running_or_queued[new_job_id] = i
# assess outcome of completed jobs
for job_id in queue_status.keys():
if queue_status[job_id] == 'C' and job_id in running_or_queued.keys():
task_id = running_or_queued[job_id][0]
command_index = running_or_queued[job_id][1]
command_dict = task_summaries[task_id]['commands'][command_index]
check_successful = determine_check_successful_function( command_dict , single_relax = single_relax )
success = check_successful( command_dict )
failure_summary = ''
if isinstance( success , bool ):
complete = success
elif len( success ) > 1 and isinstance( success[0] , bool ):
complete = success[0]
failure_summary += ' '+ ';'.join( [str( j ) for j in success[1:]] ) +' '
# track the number of attempts?
# try until failure - how many times?
tries = 0
if 'run' in command_dict.keys() and command_dict['run'] and not 'success' in command_dict['run'] and not 'failure' in command_dict['run']:
tries = int( command_dict['run'] )
tries += 1
if tries >= max_pbs_tries:
# its a failure
failure_summary = 'success'*complete + (str( tries ) +' tries;failure ' + failure_summary)*(not complete)
else:
# record the number of tries
failure_summary = str( tries )
# update the record
task_summaries[task_id]['commands'][command_index]['run'] = failure_summary
# optionally cleanup
if ddg_monomer_cleanup and command_dict['feature'] == 'ddg_monomer':#'ddg' in i['output_filename']:
print 'ddg_monomer writes useless output files, deleting these now...'
remove_intermediate_ddg_monomer_files()
# jobs that have since been completed - consider them complete?
completed.append( running_or_queued[job_id] )
del running_or_queued[job_id]
# write out "completed"? or "running_or_queued"?
# update task_summaries e.g. write them!
# modified: so the task summary records its own name...bah!
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
write_task_summary( i , i['filenames']['task_summary_filename'] )
# pause...
time.sleep( PBS_QUEUE_MONITOR_DELAY )
# return anything?
# write one last time?
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
write_task_summary( i , i['filenames']['task_summary_filename'] )
def run_VIPUR_tasks_in_batch_SLURM( task_summaries , task_list , max_slurm_tries = 2 , ddg_monomer_cleanup = True , single_relax = True ):
# also setup to do start-stop
completed = [i for i in task_list if 'run' in task_summaries[i[0]]['commands'][i[1]] and 'success' in task_summaries[i[0]]['commands'][i[1]]['run']]
attempt = 1
while not len( completed ) == len( task_list ):
# do not worry about the queue in this mode
# assume you will be able to submit etc.
# not such thing as "running or queued" either, just run one batch at a time
# gather all the commands into a single script
# choose the next job
jobs_to_run = [i for i in task_list if
not i in completed and
not ( 'run' in task_summaries[i[0]]['commands'][i[1]] and
('success' in task_summaries[i[0]]['commands'][i[1]]['run'] or
'failure' in task_summaries[i[0]]['commands'][i[1]]['run']) )
]
print str( len( jobs_to_run ) ) + ' processes still need to finish'
# write the script
master_script_text = '\n\n'.join( [task_summaries[i[0]]['commands'][i[1]]['command'] for i in jobs_to_run] )
# test without relax processes
# master_script_text = '\n\n'.join( [task_summaries[i[0]]['commands'][i[1]]['command'] for i in jobs_to_run if not 'relax' in task_summaries[i[0]]['commands'][i[1]]['command']] )
master_script_text = SLURM_BASH_SCRIPT( master_script_text )
# check if they have different names!?...wait...they will...
slurm_script_filename = task_summaries[0]['filenames']['slurm_script_filename']
slurm_output_filename = task_summaries[0]['filenames']['slurm_output_filename']
slurm_error_filename = task_summaries[0]['filenames']['slurm_error_filename']
slurm_script_filename = slurm_script_filename.replace( '.sh' , '_'+ str( attempt ) + '.sh' )
slurm_output_filename = slurm_output_filename.replace( '.out' , '_'+ str( attempt ) + '.out' )
slurm_error_filename = slurm_error_filename.replace( '.err' , '_'+ str( attempt ) + '.err' )
# can just use the first one now...
f = open( slurm_script_filename , 'w' )
f.write( master_script_text )
f.close()
# save a copy of this script for reference?
# successive runs with overwrite the file...
# debug
# raw_input( 'everything okay?' )
# submit sbatch
# simple for now...
command = 'sbatch -n 40'
if slurm_output_filename:
command += ' -o ' + slurm_output_filename
if slurm_error_filename:
command += ' -e ' + slurm_error_filename
command += ' ' + slurm_script_filename
batch_job_id = run_local_commandline( command , collect_stdout = True )
# batch_job_id = run_local_commandline( 'sbatch -n 40 ' + slurm_script_filename , collect_stdout = True )
# srun or sbatch?
batch_job_id = batch_job_id.strip().split( ' ' )[-1]
print 'submitted ' + batch_job_id
# monitor the job until it is complete
# pause...
batch_complete = False
while not batch_complete:
queue_status = get_slurm_queue_status( only_job_status = True )
# batch_complete = bool( [i for i in queue_status if i[0] == batch_job_id] )
batch_complete = not batch_job_id in queue_status.keys()
# could be an immediate failure...but don't want to linger here anyway in that case
# debug
# print queue_status
# print queue_status.keys()
# print batch_complete , batch_job_id , batch_job_id in queue_status.keys()
for i in queue_status.keys():
print i + '\t' + queue_status[i]
# can be sure it doesn't need to wait if empty
if queue_status:
print 'waiting ' + str( SLURM_QUEUE_MONITOR_DELAY ) +'s...'
time.sleep( SLURM_QUEUE_MONITOR_DELAY )
# evaluate if it ran successfully
for job_pair in jobs_to_run:
command_dict = task_summaries[job_pair[0]]['commands'][job_pair[1]]
check_successful = determine_check_successful_function( command_dict , single_relax = single_relax )
success = check_successful( command_dict )
failure_summary = ''
if isinstance( success , bool ):
complete = success
elif len( success ) > 1 and isinstance( success[0] , bool ):
complete = success[0]
failure_summary += ' '+ ';'.join( [str( j ) for j in success[1:]] ) +' '
# track the number of attempts?
# try until failure - how many times?
tries = 0
if 'run' in command_dict.keys() and command_dict['run'] and not 'success' in command_dict['run'] and not 'failure' in command_dict['run']:
tries = int( command_dict['run'] )
tries += 1
job_task = task_summaries[job_pair[0]]['root_filename'].split( '/' )[-1]
job_feature = task_summaries[job_pair[0]]['commands'][job_pair[1]]['feature']
job_variant = ''
if 'variant' in task_summaries[job_pair[0]]['commands'][job_pair[1]].keys():
job_variant = task_summaries[job_pair[0]]['commands'][job_pair[1]]['variant']
this_job_description = job_task +' '+ job_feature + (' ' + job_variant)*bool( job_variant )
if tries >= max_slurm_tries:
# its a failure
print this_job_description + ' completed successfully'*complete + (' failed with ' + str( tries ) + ' attempts')*(not complete)
failure_summary = 'success'*complete + (str( tries ) +' tries;failure ' + failure_summary)*(not complete)
completed.append( job_pair )
elif complete:
print this_job_description + ' simply completed successfully'
failure_summary = 'success' #+ str( tries ) + ' tries'
completed.append( job_pair )
else:
# record the number of tries
print this_job_description + ' completed' + ' successfully'*complete
failure_summary = str( tries )
completed.append( job_pair )
# update the record
task_summaries[job_pair[0]]['commands'][job_pair[1]]['run'] = failure_summary
# no need to be here anymore
# optionally cleanup
# if ddg_monomer_cleanup and command_dict['feature'] == 'ddg_monomer':#'ddg' in i['output_filename']:
# print 'ddg_monomer writes useless output files, deleting these now...'
# remove_intermediate_ddg_monomer_files()
# update task_summaries e.g. write them!
# modified: so the task summary records its own name...bah!
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
print 'updating: ' + i['filenames']['task_summary_filename']
write_task_summary( i , i['filenames']['task_summary_filename'] )
# debug
# print attempt
# print len( completed ) , len( task_list )
# raw_input( 'start the next round?' )
# need to run another batch?
attempt += 1
def create_variant_protein_structures(pdb_filename,
variants,
chain,
use_pyrosetta=USE_PYROSETTA,
pymol_environment_setup=''):
# optionally run environment setup
if pymol_environment_setup:
print 'setting up environment variables'
run_local_commandline(pymol_environment_setup)
# make sure the variants have been filtered
if use_pyrosetta:
# load the PDB as a pose
pose = pose_from_pdb(pdb_filename)
failed = {}
root_filename = pdb_filename.rstrip('pdb')
# currently cannot handle multi-chain input
# handle this before VIPUR
if pose.chain(pose.total_residue()) > 1:
print 'CANNOT currently handle multi-chain PDBs (as pose), using PyMOL instead!'
if PATH_TO_PYMOL:
create_variant_protein_structures(pdb_filename,
variants,
chain,
use_pyrosetta=False)
return
else:
faulty = 'clean before VIPUR, cannot handle multi-chain PDBs'
failed[faulty] = variants
elif not pose.pdb_info().chain(1) == chain:
print '...not sure what it happening, you wanted chain ' + chain + ' but VIPUR found chain ' + pose.chain(
1) + ', skipping this entire sample!'
faulty = 'clean before VIPUR, improper chain ID'
failed[faulty] = variants
# in case this condition is found:
faulty = 'could not load position from PDB'
variant_structures = []
for variation in variants:
# make a copy
test_pose = Pose()
test_pose.assign(pose)
native = variation[0]
position = variation[1:-1]
mutant = variation[-1]
# make sure the position was loaded
icode = ' ' # default...this could cause problems...
if not position[-1].isdigit():
icode = position[-1]
position = position[:-1]
position = int(position)
if not test_pose.pdb_info().pdb2pose(chain, position, icode):
if faulty in failed.keys():
failed[faulty].append(variation)
else:
failed[faulty] = [variation]
break # stop the loop
position = test_pose.pdb_info().pdb2pose(chain, position, icode)
# simple, use a mover to make the change
# appears to have trouble with N terminal variants since it uses "replace_residue"
# code is available that does not have this problem, however reloading into Rosetta with accurately determine the position of these atoms
make_variant = MutateResidue(position, mutant)
make_variant.apply(test_pose)
# write out
out_filename = self.root_filename + '.chain_' + chain + '_' + variation + '.pdb'
variant_structures.append(out_filename)
test_pose.dump_pdb(out_filename)
print 'generated ' + variation + ' variant structure and wrote to ' + out_filename
return variant_structures
else:
# use the pymol script
#for variants in self.variants['permissible']:
# use default output naming
# create command explicitly here, slightly different
root_filename = pdb_filename.rstrip('.pdb')
command = PATH_TO_PYMOL + ' -qcr ' + PATH_TO_VIPUR + '/pymol_make_variant_structure.py -- -p ' + pdb_filename + ' -m ' + ','.join(
variants) + ' -c ' + chain + ' -r ' + root_filename
if pymol_environment_setup:
command = pymol_environment_setup + '\n\n' + command
run_local_commandline(command)
# reconstruct the names
variant_structures = [
root_filename + '.chain_' + chain + '_' + i + '.pdb'
for i in variants
]
# verify they have been made
if [None for i in variant_structures if not os.path.isfile(i)]:
raise IOError(
'could not make variant protein structures,\ntry checking the input PDB file or the pymol script pymol_make_variant_structure.py'
)
return variant_structures
def run_VIPUR_tasks_SLURM( task_summaries , task_list , max_pbs_tries = 2 , ddg_monomer_cleanup = True , single_relax = False , delete_intermediate_relax_files = False ):
# run the non_rescore tasks
completed = [i for i in task_list if 'run' in task_summaries[i[0]]['commands'][i[1]] and 'success' in task_summaries[i[0]]['commands'][i[1]]['run']]
# should running_or_queued be saved? written to file?
running_or_queued = {}
rounds = 0
while not len( completed ) == len( task_list ):
rounds += 1
print '\n\nQUEUE MONITOR ROUND ' + str( rounds )
# debug
# print running_or_queued
# check queue status
queue_status = get_slurm_queue_status( only_job_status = True )
# update "running_or_queued" list (?)
# err, no, does not have information on which job it is...:(
#for i in queue_status.keys():
# if queue_status[i] in ['R' , 'Q']:
# used to be after submission, not occurs first
# debug, need to know
running_jobs = len( [i for i in queue_status.values() if i in ['R']] )
if running_jobs:
print str( running_jobs ) + ' are still running...'
# assess outcome of completed jobs
# still_running = 0
# need to add the jobs that completed, removed themselves from the queue in SLURM
# print queue_status.keys() + [j for j in running_or_queued.keys() if not j in queue_status.keys()]
for job_id in queue_status.keys() + [j for j in running_or_queued.keys() if not j in queue_status.keys()]:
# debug
# if job_id in queue_status.keys():
# print '\t'+ job_id , queue_status[job_id] , job_id in running_or_queued.keys()
# else:
## print '\t'+ job_id , None , job_id in running_or_queued.keys()
# print '\t'+ job_id , job_id in running_or_queued.keys()
if (not job_id in queue_status.keys()) or (queue_status[job_id] == 'C' and job_id in running_or_queued.keys()):
task_id = running_or_queued[job_id][0]
command_index = running_or_queued[job_id][1]
command_dict = task_summaries[task_id]['commands'][command_index]
check_successful = determine_check_successful_function( command_dict , single_relax = single_relax )
success = check_successful( command_dict )
failure_summary = ''
if isinstance( success , bool ):
complete = success
elif len( success ) > 1 and isinstance( success[0] , bool ):
complete = success[0]
failure_summary += ' '+ ';'.join( [str( j ) for j in success[1:]] ) +' '
# track the number of attempts?
# try until failure - how many times?
tries = 0
if 'run' in command_dict.keys() and command_dict['run'] and not 'success' in command_dict['run'] and not 'failure' in command_dict['run']:
tries = int( command_dict['run'] )
tries += 1
if tries >= max_pbs_tries:
# its a failure
print job_id + ' completed successfully'*complete + (' failed with ' + str( tries ) + ' attempts')*(not complete)
failure_summary = 'success'*complete + (str( tries ) +' tries;failure ' + failure_summary)*(not complete)
elif complete:
print job_id + ' simply completed successfully'
failure_summary = 'success' #+ str( tries ) + ' tries'
else:
# record the number of tries
print job_id + ' completed' + ' successfully'*complete
failure_summary = str( tries )
# update the record
task_summaries[task_id]['commands'][command_index]['run'] = failure_summary
# optionally cleanup
if ddg_monomer_cleanup and command_dict['feature'] == 'ddg_monomer':#'ddg' in i['output_filename']:
print 'ddg_monomer writes useless output files, deleting these now...'
remove_intermediate_ddg_monomer_files()
# jobs that have since been completed - consider them complete?
completed.append( running_or_queued[job_id] )
del running_or_queued[job_id]
# write out "completed"? or "running_or_queued"?
# else:
# still_running += 1
# print str( still_running) + ' jobs still running (or queued)...'
# update task_summaries e.g. write them!
# modified: so the task summary records its own name...bah!
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
print 'updating: ' + i['filenames']['task_summary_filename']
write_task_summary( i , i['filenames']['task_summary_filename'] )
# used to be first, submit jobs them check complete
# but slurm removes jobs from the list
queue_space_occupied = len( [i for i in queue_status.values() if not i in ['C' , 'PD']] ) # ignore "C"ompleted jobs, "R"unning job quota are not set by us...
# if your queue system does not have a separate "R"un quota, remove 'R' from the above!
available_space = SLURM_QUEUE_QUOTA - queue_space_occupied
# launch next jobs in available slots
if available_space:
print str( queue_space_occupied ) + ' jobs queued or running, could submit up to ' + str( available_space ) + ' more'
# choose the next job
jobs_to_run = [i for i in task_list if
not i in completed and
not i in running_or_queued.values() and
not ( 'run' in task_summaries[i[0]]['commands'][i[1]] and
('success' in task_summaries[i[0]]['commands'][i[1]]['run'] or
'failure' in task_summaries[i[0]]['commands'][i[1]]['run']) )
]
print str( len( jobs_to_run ) ) + ' jobs still need to finish (after the currently running jobs complete)'
# only the next few
for i in jobs_to_run[:available_space]:
command_dict = task_summaries[i[0]]['commands'][i[1]]
# write scripts as part of pre processing?...yeah...
# write the command to a script
#script_filename = command_dict['out_path'] +'/'*bool( command_dict['out_path'] )+
#script_filename = command_dict['script_filename']
# submission is specific to the job
slurm_command = ''
# if its a rescore and relax jobs were separated, need to recombine them!
if 'rescore' in command_dict['feature']:
# combine the individual relax runs
#relax_commands = [i for i in task_summary['commands'] if i['feature'].replace( '_native' , '' ) == 'relax']
#silent_filenames = [j['output_filename'] for j in relax_commands if j['variant'] == i['variant'] and 'run' in j.keys() and j['run'] == 'success']
silent_filenames = [j['output_filename'] for j in task_summaries[i[0]]['commands'] if
j['feature'].replace( '_native' , '' ) == 'relax' and
j['variant'] == command_dict['variant'] and
'run' in j.keys() and
'success' in j['run']
]
# actually need to identify the combined_silent_filename, be sure the relax files have not already been merged
# which variant
target_variant = [j for j in task_summaries[i[0]]['variants'].keys() if j.split( '_' )[-1] == command_dict['variant'] and j.split( '_' )[0] in command_dict['command']]
if not target_variant:
# its native
combined_silent_filename = task_summaries[i[0]]['other']['combined_native_silent_filename']
combined_score_filename = task_summaries[i[0]]['other']['combined_native_score_filename']
elif len( target_variant ) > 1:
raise Exception( '??? found more than on matching variant ???\n' + ', '.join( target_variant ) )
else:
# found it
combined_silent_filename = task_summaries[i[0]]['variants'][target_variant[0]]['combined_silent_filename']
combined_score_filename = task_summaries[i[0]]['variants'][target_variant[0]]['combined_score_filename']
#if not single_relax: # AND post processing has not already be run...scan for the combined silent file
if not single_relax and not os.path.isfile( combined_silent_filename ):
if not len( silent_filenames ) == ROSETTA_RELAX_OPTIONS['nstruct']:
raise Exception( '??? somehow the matching relax run(s) has failed ???\n' + str( i ) )
score_filenames = [j.replace( '.silent' , '.sc' ) for j in silent_filenames]
merge_rosetta_relax_output( silent_filenames , combined_silent_filename , score_filenames , combined_score_filename , delete_old_files = delete_intermediate_relax_files )
# rescore already knows the proper filename
else:
# just a single match for each
# output filename should be correct as is :)
None
# do this above, in initial processing
# elif 'ddg_monomer' in command_dict['feature']:
# must do it this way for now
# write the run script
# ddg_monomer_script_filename = task_summary['filenames']['slurm_script_filename'].replace( 'slurm_script_this_batch.sh' , 'run_ddg_momomer_script.sh' )
# f = open( ddg_monomer_script_filename , 'w' )
# f.write( command_dict['command'] )
# f.close()
# submit this script using a queue command
# srun or sbatch?
slurm_command = command_dict['sbatch_command'] # SHOULD already have an abspath to the script
new_job_id = run_local_commandline( slurm_command , collect_stdout = True )
# new_job_id = new_job_id[:new_job_id.find( ' ' )]
new_job_id = new_job_id.strip().split( ' ' )[-1]
print 'submitted ' + new_job_id
# save the job id
# assume its queue
running_or_queued[new_job_id] = i
# print 'added ' + new_job_id
else:
print 'no new \"positions\" are available'
# OKAY, move the "updating" to just after the status check
# problem with ddg_monomer, runs so fast...
# make a specific exception:
# ...move the pause to here
# prevents odd behaviour...um...sorta? maybe not
# if 'ddg_monomer' in command_dict['feature']:
# used to be where the updates were compared
# pause...
if queue_space_occupied or jobs_to_run:
print 'waiting ' + str( SLURM_QUEUE_MONITOR_DELAY ) +'s...'
time.sleep( SLURM_QUEUE_MONITOR_DELAY )
# return anything?
# write one last time?
for i in task_summaries:
if not 'task_summary_filename' in i['filenames'].keys():
raise NotImplementedError( 'should input the task summary filename (not the summary itself)...' )
else:
# write it out
write_task_summary( i , i['filenames']['task_summary_filename'] )