def calculateInteractions(job,
localExe=True,
killer=None,
sshConn=None,
dock_soft='PLANTS',
timemd=2.5,
clean=False,
remoteSettings=None,
modelFN='model.dat'):
logging.info("Calculating interactions for job: %s" % job.filename)
logging.info("State job %s: %s" % (job.filename, job.status))
if job.status == 'COMPLETED':
return
if killer == None:
killer = jobHandler.WarmKiller()
killSig = os.path.join(job.dirTemp, 'KILL')
initialStatus = job.status
#LOAD MODEL INFO
success, model = loadModel(os.path.join(modelDir, job.modelProt))
if not success:
job.status = 'FAILED'
job.results = model
# START CHECK/OPERATIONS
try:
if not os.path.isdir(job.dirTemp):
os.mkdir(job.dirTemp)
except Exception, e:
job.status = 'FAILED'
job.results = "Working directory not found or its creation failed: %s " % (
e)
logging.error('{0} {1}'.format(job.status, job.results))
def modelDesc(prot):
protDir = os.path.join(settings.get('etoxlie_model_dir'), prot)
success, model = modelHandler.loadModel(protDir,
prediction=False,
verModel=0,
modelFN='model.dat')
if not success:
return "Model not found :'-("
logging.debug('Load model {0} version {1} from: {2}'.format(
prot, 0, protDir))
# Source model parameters
return render_template('modelDesc.html', model=model)
def modelDescVer(prot, ver):
protDir = os.path.join(settings.get('etoxlie_model_dir'), prot)
success, (model, params) = modelHandler.loadModel(protDir,
prediction=True,
verModel=int(ver),
modelFN='model.dat')
logging.debug('Load model {0} version {1} from: {2}'.format(
prot, ver, protDir))
# compute pearson r, spearman s
Gcalc = [float(x['Gcalc']) for x in params['trainSet']]
Gexp = [float(x['Gexp']) for x in params['trainSet']]
idxs = sorted(list(range(len(Gcalc))), key=lambda x: Gcalc[x])
rankGcalc = [0] * len(idxs)
for i, x in enumerate(idxs):
rankGcalc[x] = i
idxs = sorted(list(range(len(Gexp))), key=lambda x: Gexp[x])
rankGexp = [0] * len(idxs)
for i, x in enumerate(idxs):
rankGexp[x] = i
s = corrcoef(rankGexp, rankGcalc)[0, 1]
r = corrcoef(Gexp, Gcalc)[0, 1]
params['r'] = r
params['s'] = s
params['date'] = time.strftime("%Y/%m/%d\n%H:%M:%S",
time.localtime(params['creation_date']))
img = plotting.plotLIE(params['trainSet'])
svg = plotting.fixSVG(img, h=400)
return render_template('modelVersDesc.html', param=params,
plot=svg) #send_file(a, mimetype='image/svg+xml'))
def submitScreen(sdfFn,
modelId,
prediction,
etoxlie_folder='.',
fieldExp='Activity',
misc=None,
jobid=None):
# sdfFn: filename sdf
# modelId: {'modelProt': 1A2; 'modelProtVer':1}
# prediction: True or False
# in case of prediction=False: misc={'isGamma':True|False,'fixBeta':True|False}
# 1. check sdf
status = 'SUBMITTED'
if os.path.exists(sdfFn):
with open(sdfFn, 'r') as infile:
sdf = infile.read()
else:
results = "File not found: %s" % sdfFn
status = 'FAILED'
sdf = ''
# 2 Checking consistency of work and models
if status != 'FAILED':
# A. check other runs, in case start
listJobs = jobHandler.collectJobs(cpd=False, sort=True)
for job in listJobs:
if job.status != 'FAILED':
if sdf == job.sdf:
if modelId['modelProt'] == job.modelProt:
if modelId['modelProtVer'] == job.modelProtVer:
if prediction == job.prediction:
if not prediction:
if misc == job.experiment:
status = 'FAILED'
results = 'SDF screen is already in process (jobID: %s)' % job.filename
if status != 'FAILED':
# B. check model in case quit
success, results = loadModel(os.path.join(modelDir,
modelId['modelProt']),
prediction=prediction,
verModel=modelId['modelProtVer'])
if success:
if prediction:
model, params = results
else:
model = results
else:
results = results
status = 'FAILED'
# 3 Create job
if status != 'FAILED':
jobSDR = jobHandler.jobCpd()
jobSDR.create(sdf,
modelId['modelProt'],
modelId['modelProtVer'],
prediction=prediction,
jobid=jobid)
jobSDR.sdf = sdf
jobSDR.experiment = misc
# 4 Now submit jobs (if already submitted, link to the existing
if status != 'FAILED':
success, results = submitCPD(sdf, modelId['modelProt'],
modelId['modelProtVer'], prediction,
fieldExp)
if success:
jobSDR.results = results
else:
jobSDR.status = 'FAILED'
jobSDR.results = results
jobSDR.update()
if status == 'FAILED':
return (False, results)
else:
return (True, jobSDR.filename)
def screenSDF(listCpds, prediction, modelProt, modelProtVer, misc):
# Check all the processes of a sdf screening, update status job and status single cpds runs in results
updated = False
status = 'SUBMITTED'
try:
completed = True
for nc, cpd in enumerate(listCpds):
# Check status single cpd (only if not already finished)
if cpd['Status'] not in statEndJob:
completed = False
jobCpd = jobHandler.jobCpd()
jobCpd.load(os.path.join(modelDir, cpd['JobName']))
#Update progress
if cpd['Status'] != jobCpd.status:
updated = True
listCpds[nc]['Status'] = jobCpd.status
listCpds[nc]['Results'] = jobCpd.results
if jobCpd.status == 'DONE':
if prediction:
### PREDICT COMPOUND AND UPDATE RESULTS WITH PREDICTION
#here function for pred and AD. Results is a list of dictionaries containing energies for each simulation
data = getEnergies([listCpds[nc]])
success, results = loadModel(os.path.join(
modelDir, modelProt),
prediction=True,
verModel=modelProtVer)
if not success:
print results
listCpds[nc]['Status'] = 'FAILED'
#raise Exception, results
continue
else:
modelLIE, params = results
try:
Gcalc, idposes, wi, sdep, sspred = predictLie(
data, params['LIE'])
listCpds[nc]
listCpds[nc]['DGexp'] = None
listCpds[nc]['DGcalc'] = Gcalc[0]
listCpds[nc]['idpose'] = idposes[0]
listCpds[nc]['wi'] = wi[0]
CI = predictAD(data, {
'DGcalc': Gcalc,
'idposes': idposes,
'wi': wi
}, params['AD'])
listCpds[nc]['CI_analysis'] = CI[0]
listCpds[nc]['CI'] = sum(CI[0].values())
listCpds[nc]['Err'] = predictError(
sum(CI[0].values()), params['LIE']['sdep'])
logging.info('RESULTS:')
for item in listCpds[nc]:
logging.info("{0} {1}".format(
item, listCpds[nc][item]))
logging.info('END RESULTS')
except Exception, e:
logging.error(
'Error in computing energy for job %s: %s'
% (cpd['JobName'], e))
listCpds[nc]['Status'] = 'FAILED'
continue
if completed:
logging.info('Job completed')
if not prediction:
# check type of calibration for backward compatibility:
if misc == None:
misc = {'isGamma': False, 'fixBeta': False}
# Here collect energies
print 'GET ENERGIES'
data = getEnergies(
listCpds
) # results is a dictionary with {ids, Y, ene, smi, decVdw, decEle}
logging.info(
'Starting calibration for protein %s, version %s' %
(modelProt, modelProtVer))
# Here create model
print 'CALIBRATE LIE'
LIEmodel = calibrateLie(data,
gamma=misc['isGamma'],
fixBeta=misc['fixBeta'])
print 'CALIBRATE AD'
ADmodel = calibrateAD(data, LIEmodel)
modelId = {
'modelProt': modelProt,
'modelProtVer': '%04d' % modelProtVer
}
print 'SAVE MODEL'
success, out = saveModel(modelDir, modelId, LIEmodel, ADmodel)
if not success:
status = 'FAILED'
raise Exception, out
logging.info('New model saved on file %s' % out)
deltaId = 0
for n, id in enumerate(data['ids']):
if listCpds[id + deltaId]['Status'] == 'DONE':
listCpds[id + deltaId]['DGcalc'] = LIEmodel['Gcalc'][n]
listCpds[id + deltaId]['idpose'] = LIEmodel['idx'][n]
listCpds[id + deltaId]['wi'] = LIEmodel['wi'][n]
else:
deltaId += 1
status = 'DONE'
updated = True
results = listCpds