def loadResults_Mitocheck(self,exp_list):
'''
Here we're loading results on a per experiment basis. This will be interesting to look at distances between experiments
based on phenotypes, vs distances based on trajectory types.
'''
if len(exp_list[0])!=2:
exp_list=strToTuple(exp_list, os.listdir(self.settings.outputFolder))
result = None; i=0; missed=[]
for pl,w in exp_list:
print i,
try:
f=open(os.path.join(self.settings.outputFolder,pl, self.settings.outputFile.format(pl[:10], w)), 'r')
pheno_seq_list, mask = pickle.load(f)
f.close()
except:
print "Loading error for ", pl, w
missed.append(i)
continue
else:
pheno_seq_list = np.sum( np.array([np.bincount(pheno_seq_list[j], minlength=17) for j in range(len(pheno_seq_list)) if j not in mask]), 0)[:-2]
#15 and 16 are respectively out of focus and artefact objects. We don't want them
pheno_seq_list=pheno_seq_list/float(np.sum(pheno_seq_list))
result = np.vstack((result, pheno_seq_list)) if result is not None else pheno_seq_list
finally:
i+=1
print "Saving"
f=open(os.path.join(self.settings.outputFolder,self.settings.outputFile.format("ALL", "hit_exp")), 'w')
pickle.dump((result, missed),f); f.close()
return
def scriptCommand(exp_list, baseName='comp_track', command="tracking/trajPack/cell_cycle.py",jobSize=10,
h5_result_dir="/share/data20T/mitocheck/Alice/results",
max_nodes=500,
**kwargs):
perExperiment=False
if type(exp_list[0])!=int:
perExperiment=True
if len(exp_list[0])!=2:
exp_list=strToTuple(exp_list, os.listdir(h5_result_dir))
fileNumber = int(len(exp_list)/float(jobSize))+1
head = """#!/bin/sh
cd %s""" %progFolder
for keyword in kwargs:
command = command +" -{} {}".format(keyword, kwargs[keyword])
for k in range(fileNumber):
cmd = ''
for exp in exp_list[jobSize*k:jobSize*(k+1)]:
if perExperiment:
pl,w=exp
temp_cmd = """
python %s -p %s -w %s"""
temp_cmd %= (
command,
pl,
w
)
else:
temp_cmd = """
python %s -i %i"""
temp_cmd %= (
command,
exp
)
cmd += temp_cmd
# this is now written to a script file (simple text file)
# the script file is called ltarray<x>.sh, where x is 1, 2, 3, 4, ... and corresponds to the job index.
script_name = os.path.join(scriptFolder, '%s%i.sh' % (baseName, k+1))
script_file = file(script_name, "w")
script_file.write(head + cmd)
script_file.close()
# make the script executable (without this, the cluster node cannot call it)
os.system('chmod a+x %s' % script_name)
# write the main script
array_script_name = '%s.sh' % os.path.join(scriptFolder, baseName)
main_script_file = file(array_script_name, 'w')
main_content = """#!/bin/sh
%s
#$ -o %s
#$ -e %s
%s$%s.sh
""" % (path_command,
pbsOutDir,
pbsErrDir,
os.path.join(scriptFolder, baseName),
pbsArrayEnvVar)
main_script_file.write(main_content)
os.system('chmod a+x %s' % array_script_name)
sub_cmd = 'qsub -tc %i -t 1-%i %s' % (max_nodes, fileNumber, array_script_name)
print sub_cmd
return 1
def globalSummaryScript(baseName, siRNAFile,
n_clusters_min, n_clusters_max,
div_name, lambda_, weights,
bins_type, bin_size, cost_type,
batch_size, n_init, init,
ddim, iter_=0):
f=open(siRNAFile, 'r')
siRNAList = pickle.load(f); f.close()
siExpDict = expSi(qc = quality_control_file , sens=0)
jobCount = 0
i=0
total_expList = []
head = """#!/bin/sh
cd %s""" %progFolder
baseName = baseName+'{}{}_w{}_{}_{}_{}'.format(iter_,div_name[:5], weights, bins_type, bin_size, cost_type)
#A. DEALING WITH EXPERIMENTS
for siRNA in siRNAList:
try:
expList = siExpDict[siRNA]
except KeyError:
print "siRNA not in siRNA-experiment dictionary"
else:
expList = strToTuple(expList, os.listdir(data_folder))
total_expList.extend(expList)
for plate, well in expList:
jobCount += 1; i+=1
cmd = plateWellSummaryScript(plate, well, div_name, lambda_, weights, bins_type, bin_size, cost_type, batch_size, n_init, init, ddim, iter_)
# this is now written to a script file (simple text file)
# the script file is called ltarray<x>.sh, where x is 1, 2, 3, 4, ... and corresponds to the job index.
script_name = os.path.join(scriptFolder, baseName+'{}.sh'.format(i))
script_file = file(script_name, "w")
script_file.write(head + cmd)
script_file.close()
# make the script executable (without this, the cluster node cannot call it)
os.system('chmod a+x %s' % script_name)
#B. DEALING WITH CONTROLS
ctrlExp = appendingControl(total_expList)
ctrlExp = countingDone(ctrlExp)
np.random.shuffle(ctrlExp)
ctrlExp=ctrlExp[:int(0.2*len(total_expList))]
for plate, well in ctrlExp:
jobCount += 1; i+=1
cmd = plateWellSummaryScript(plate, well, div_name, lambda_, weights, bins_type, bin_size, cost_type, batch_size, n_init, init, ddim, iter_)
# this is now written to a script file (simple text file)
# the script file is called ltarray<x>.sh, where x is 1, 2, 3, 4, ... and corresponds to the job index.
script_name = os.path.join(scriptFolder, baseName+'{}.sh'.format(i))
script_file = file(script_name, "w")
script_file.write(head + cmd)
script_file.close()
# make the script executable (without this, the cluster node cannot call it)
os.system('chmod a+x %s' % script_name)
# write the main script
array_script_name = '%s.sh' % os.path.join(scriptFolder, baseName)
main_script_file = file(array_script_name, 'w')
main_content = """#!/bin/sh
%s
#$ -o %s
#$ -e %s
%s$%s.sh
""" % (path_command,
pbsOutDir,
pbsErrDir,
os.path.join(scriptFolder, baseName),
pbsArrayEnvVar)
main_script_file.write(main_content)
main_script_file.close()
os.system('chmod a+x %s' % array_script_name)
sub_cmd = 'qsub -t 1-%i %s' % (jobCount, array_script_name)
print sub_cmd
#C. DOING EXPERIMENT CLUSTERING STEP
expFilename = 'exp_Simpson_{}.pkl'.format(int(time.time()))
total_expList.extend(ctrlExp)
f=open(expFilename, 'w')
pickle.dump(total_expList, f)
f.close()
baseName = baseName+'_clustering'
for n_clusters in range(n_clusters_min, n_clusters_max):
script_name = os.path.join(scriptFolder, baseName+'{}.sh'.format(n_clusters-n_clusters_min))
script_file = file(script_name, "w")
cmd="""
python tracking/histograms/summarization_clustering.py -a clustering --experimentFile %s -k %i --ddimensional %i --bins_type %s --cost_type %s --bin_size %i --div_name %s -w %i --init %s --batch_size %i --iter %i
"""
cmd %= (
expFilename,
n_clusters,
ddim,
bins_type,
cost_type,
bin_size,
div_name,
weights,
init,
batch_size,
iter_
)
script_file.write(head + cmd)
script_file.close()
os.system('chmod a+x %s' % script_name)
# write the main script
array_script_name = '%s.sh' % os.path.join(scriptFolder, baseName)
main_script_file = file(array_script_name, 'w')
main_content = """#!/bin/sh
%s
#$ -o %s
#$ -e %s
%s$%s.sh
""" % (path_command,
pbsOutDir,
pbsErrDir,
os.path.join(scriptFolder, baseName),
pbsArrayEnvVar)
main_script_file.write(main_content)
main_script_file.close()
os.system('chmod a+x %s' % array_script_name)
sub_cmd = 'qsub -hold_jid -t 1-%i %s' % (n_clusters_max - n_clusters_min, array_script_name)
print sub_cmd
#D. GOING BACK TO EXPERIMENTS AND TESTING IF DIFFERENT FROM CONTROLS
return 1
def hitFinderScript(baseName, siRNAFile, testCtrl=False, iter_=0):
f=open(siRNAFile, 'r')
siRNAList = pickle.load(f); f.close()
jobCount = 0
head = """#!/bin/sh
cd %s""" %progFolder
baseName+='{}'.format(iter_)
if testCtrl:
baseName = baseName+'CTRL'
expList = []
yqualDict=expSi(quality_control_file, sens=0)
for siRNA in siRNAList:
expList.extend(strToTuple(yqualDict[siRNA], os.listdir(data_folder)))
plates = Counter(np.array(expList)[:,0]).keys()
for i,plate in enumerate(plates):
jobCount+=1; i+=1
cmd = '''
python tracking/histograms/summarization_clustering.py -a hitFinder --verbose 0 --testCtrl %s --iter %i
'''
cmd%=(plate, iter_)
# this is now written to a script file (simple text file)
# the script file is called ltarray<x>.sh, where x is 1, 2, 3, 4, ... and corresponds to the job index.
script_name = os.path.join(scriptFolder, baseName+'{}.sh'.format(i))
script_file = file(script_name, "w")
script_file.write(head + cmd)
script_file.close()
# make the script executable (without this, the cluster node cannot call it)
os.system('chmod a+x %s' % script_name)
else:
for i,siRNA in enumerate(siRNAList):
jobCount+=1; i+=1
cmd = '''
python tracking/histograms/summarization_clustering.py -a hitFinder --siRNA %s --verbose 0 --iter %i
'''
cmd%=(siRNA, iter_)
# this is now written to a script file (simple text file)
# the script file is called ltarray<x>.sh, where x is 1, 2, 3, 4, ... and corresponds to the job index.
script_name = os.path.join(scriptFolder, baseName+'{}.sh'.format(i))
script_file = file(script_name, "w")
script_file.write(head + cmd)
script_file.close()
# make the script executable (without this, the cluster node cannot call it)
os.system('chmod a+x %s' % script_name)
# write the main script
array_script_name = '%s.sh' % os.path.join(scriptFolder, baseName)
main_script_file = file(array_script_name, 'w')
main_content = """#!/bin/sh
%s
#$ -o %s
#$ -e %s
%s$%s.sh
""" % (path_command,
pbsOutDir,
pbsErrDir,
os.path.join(scriptFolder, baseName),
pbsArrayEnvVar)
main_script_file.write(main_content)
main_script_file.close()
os.system('chmod a+x %s' % array_script_name)
sub_cmd = 'qsub -t 1-%i %s' % (jobCount, array_script_name)
print sub_cmd
return
