def run_spaan(inputFasta, outputDir, rawFlag):
command = SPAAN_PATH
rawOutput = os.path.join(outputDir, "SPAAN")
if not os.path.exists(rawOutput):
os.mkdir(rawOutput)
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
os.system(
"%s %s %s" %
(command, inputFasta,
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem)))
output = ['ID\tSPAAN_Score']
values = {}
for row in open(
os.path.join(rawOutput, "%s.output" %
Path(inputFasta).stem)).read().split('\n')[1:]:
if row == '':
continue
tokens = row.split('\t')
values[tokens[2].strip('>')] = tokens[1]
for fastaID in sequenceIDs:
if fastaID in values.keys():
output.append('\t'.join([fastaID, values[fastaID]]))
else:
output.append(fastaID + '\t0.0')
open(os.path.join(rawOutput, "%s.spaan.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
if not rawFlag:
os.remove(
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem))
def run_psortb(inputFasta, outputDir, organism, multiFlag, process,
rawFlag):
command = PSORTB_PATH
rawOutput = os.path.join(outputDir, "PSORTB")
if not os.path.exists(rawOutput):
os.mkdir(rawOutput)
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
for file in os.listdir(rawOutput):
if 'psortb_grampos.txt' in file:
os.remove(os.path.join(rawOutput, file))
if organism.lower() in ["gram+", "g+"]:
os.system("%s -p -i %s" % (command, inputFasta))
for file in os.listdir(rawOutput):
if 'psortb_grampos.txt' in file:
rawFile = file
break
elif organism.lower() in ["gram-", "g-"]:
os.system("%s -n -i %s" % (command, inputFasta))
for file in os.listdir(rawOutput):
if 'psortb_gramneg.txt' in file:
rawFile = file
break
output = [
'\t'.join([
"ID", "SubcellularLocation", "Extracellular_Probability",
"CytoplasmicMembrane_Probability", "Cytoplasmic_Probability",
"Cellwall_Probability", "Periplasmic_Probability",
"OuterMembrane_Probability"
])
]
locs = [
"Extracellular", "CytoplasmicMembrane", "Cytoplasmic", "Cellwall",
"Periplasmic", "OuterMembrane"
]
values = {}
for entry in re.split('[-]{79}',
open(os.path.join(rawOutput,
rawFile)).read())[:-1]:
fastaID = entry.strip().split('\n')[0][7:].strip()
value = [""] + ["0.0"] * (len(locs))
value[0] = re.split(
'[ ]+',
entry[entry.find("Final Prediction:"):].split('\n')[1])[1]
for row in entry[entry.find("Localization Scores:"):entry.
find("Final Prediction:")].split('\n')[1:-1]:
tokens = re.split('[ ]+', row)
value[locs.index(tokens[1]) + 1] = str(float(tokens[2]) / 10.0)
values[fastaID] = value
for fastaID in sequenceIDs:
output.append('\t'.join([fastaID] + values[fastaID]))
open(os.path.join(rawOutput, "%s.psortb.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
if not rawFlag:
os.remove(os.path.join(rawOutput, rawFile))
def split_files(inputFasta, tmpDir):
fasta = readfasta.readFasta(inputFasta, key="full", strip=False)
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
inFiles = []
outFiles = []
split = int(math.ceil(len(sequenceIDs)) / SPLIT_LIMIT)
size = int(math.ceil(len(sequenceIDs) / float(split + 1)))
for i in range(split + 1):
textBuffer = ''
for fastaID in list(fasta.keys())[i * size:(i + 1) * size]:
textBuffer = textBuffer + fastaID + fasta[fastaID]
inFile = os.path.join(tmpDir, "input.fasta.%i" % i)
open(inFile, 'w').write(textBuffer)
inFiles.append(inFile)
outFile = os.path.join(tmpDir, "output.raw.%i" % i)
outFiles.append(outFile)
return (inFiles, outFiles)
def run_immugen(inputFasta, outputDir, rawFlag):
command = IMGEN_PATH
rawOutput = os.path.join(outputDir, "IMGEN")
if not os.path.exists(rawOutput):
os.mkdir(rawOutput)
fasta = readfasta.readFasta(inputFasta, key="full")
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
tmpInput = os.path.join(rawOutput, "sequence.tmp")
open(tmpInput, 'w').write('')
for seq in fasta.values():
open(tmpInput, 'a').write("%s\n" % seq)
os.system(
"python2.7 %s %s > %s" %
(command, tmpInput,
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem)))
output = ['ID\tImmunogenicity_Score']
values = {}
for row in open(
os.path.join(rawOutput, "%s.output" %
Path(inputFasta).stem)).read().split('\n')[4:]:
if len(row) == 0:
continue
tokens = row.split(',')
seq = tokens[0]
fastaID = list(fasta.keys())[list(fasta.values()).index(seq)]
values[fastaID] = tokens[2]
for fastaID in sequenceIDs:
if fastaID in values.keys():
output.append('\t'.join([fastaID, values[fastaID]]))
else:
output.append(fastaID + '\t0.0')
open(os.path.join(rawOutput, "%s.imgen.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
if not rawFlag:
os.remove(
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem))
os.remove(tmpInput)
def makeInput(self, inputFasta, outputDir, organism, incFeatures):
featureDir = os.path.join(outputDir, "_FEATURE")
if organism.lower() in ["gram+", "g+", "gram-", "g-"]:
masterLabels = ["ID", "Gram"]
masterData = {}
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
if organism.lower() in ["gram+", "g+"]:
for fastaID in sequenceIDs:
masterData[fastaID] = [fastaID, "1"]
for method in incFeatures:
tsvFile = os.path.join(
featureDir, method.upper(),
"%s.%s.tsv" % (Path(inputFasta).stem, method))
for (i,
line) in enumerate(open(tsvFile).read().splitlines()):
tokens = line.split('\t')
if i == 0:
if method == "psortb":
masterLabels += tokens[2:]
else:
masterLabels += tokens[1:]
else:
if method == "psortb":
masterData[tokens[0]] += tokens[2:]
else:
masterData[tokens[0]] += tokens[1:]
elif organism.lower() in ["gram-", "g-"]:
for fastaID in sequenceIDs:
masterData[fastaID] = [fastaID, "0"]
for method in incFeatures:
tsvFile = os.path.join(
featureDir, method.upper(),
"%s.%s.tsv" % (Path(inputFasta).stem, method))
for (i,
line) in enumerate(open(tsvFile).read().splitlines()):
tokens = line.split('\t')
if i == 0:
if method == "psortb":
masterLabels += tokens[2:]
else:
masterLabels += tokens[1:]
else:
if method == "psortb":
masterData[tokens[0]] += tokens[2:]
else:
masterData[tokens[0]] += tokens[1:]
else:
masterLabels = ["ID"]
masterData = {}
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
for fastaID in sequenceIDs:
masterData[fastaID] = [fastaID]
for method in incFeatures:
tsvFile = os.path.join(
featureDir, method.upper(),
"%s.%s.tsv" % (Path(inputFasta).stem, method))
for (i, line) in enumerate(open(tsvFile).read().splitlines()):
tokens = line.split('\t')
if i == 0:
if method == "psortb":
masterLabels += tokens[2:]
else:
masterLabels += tokens[1:]
else:
if method == "psortb":
masterData[tokens[0]] += tokens[2:]
else:
masterData[tokens[0]] += tokens[1:]
output = ["\t".join(masterLabels)]
for fastaID in masterData.keys():
output.append("\t".join(masterData[fastaID]))
open(os.path.join(outputDir, "%s.input.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
def run_descriptor(inputFasta, outputDir, rawFlag):
rawOutput = os.path.join(outputDir, "MDESC")
if not os.path.exists(rawOutput):
os.mkdir(rawOutput)
fasta = readfasta.readFasta(inputFasta, key="full")
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
aacomp = []
ctd = []
seqord = []
autocor = []
output = []
for fastaID in sequenceIDs:
sequence = fasta[fastaID]
Des = GetProDes(sequence)
features = {}
dict.update(features, Des.GetAAComp())
keys = list(features.keys())
if len(aacomp) == 0:
aacomp.append('\t'.join(['ID'] + keys))
output.append('\t'.join(['ID'] + keys))
tmp = []
for key in keys:
tmp.append(str(features[key]))
aacomp.append('\t'.join([fastaID] + tmp))
output.append('\t'.join([fastaID] + tmp))
features = Des.GetCTD()
keys = list(features.keys())
if len(ctd) == 0:
ctd.append('\t'.join(['ID'] + keys))
output[0] += '\t%s' % '\t'.join(keys)
tmp = []
for key in keys:
tmp.append(str(features[key]))
ctd.append('\t'.join([fastaID] + tmp))
output[-1] += '\t%s' % '\t'.join(tmp)
features = {}
dict.update(features, Des.GetQSO(maxlag=MIN_PEPTIDE_LENGTH))
keys = list(features.keys())
if len(seqord) == 0:
seqord.append('\t'.join(['ID'] + keys))
output[0] += '\t%s' % '\t'.join(keys)
tmp = []
for key in keys:
tmp.append(str(features[key]))
seqord.append('\t'.join([fastaID] + tmp))
output[-1] += '\t%s' % '\t'.join(tmp)
features = {}
dict.update(features,
Des.GetMoreauBrotoAuto(maxlag=MIN_PEPTIDE_LENGTH))
dict.update(features, Des.GetGearyAuto(maxlag=MIN_PEPTIDE_LENGTH))
keys = list(features.keys())
if len(autocor) == 0:
autocor.append('\t'.join(['ID'] + keys))
output[0] += '\t%s' % '\t'.join(keys)
tmp = []
for key in keys:
tmp.append(str(features[key]))
autocor.append('\t'.join([fastaID] + tmp))
output[-1] += '\t%s' % '\t'.join(tmp)
if rawFlag:
open(
os.path.join(rawOutput,
"%s.aacomp.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(aacomp))
open(os.path.join(rawOutput, "%s.ctd.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(ctd))
open(
os.path.join(rawOutput,
"%s.seqord.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(seqord))
open(
os.path.join(rawOutput,
"%s.autocor.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(autocor))
open(os.path.join(rawOutput, "%s.mdesc.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
for fastaID in sequenceIDs:
sequence = fasta[fastaID]
def run_tmhmm(inputFasta, outputDir, multiFlag, process, rawFlag):
command = TMHMM_PATH
rawOutput = os.path.join(outputDir, "TMHMM")
if not os.path.exists(rawOutput):
os.mkdir(rawOutput)
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
if multiFlag.lower() in ['t', 'true'
] and len(sequenceIDs) > SPLIT_LIMIT:
(inFiles, outFiles) = Feature.split_files(inputFasta, rawOutput)
else:
inFiles = [inputFasta]
outFiles = [
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem)
]
clines = []
for (f, inFile) in enumerate(inFiles):
outFile = outFiles[f]
cline = "%s %s > %s" % (command, inFile, outFile)
clines.append(cline)
Feature.mp_run(clines, process)
if multiFlag.lower() in ['t', 'true'
] and len(sequenceIDs) > SPLIT_LIMIT:
Feature.combine_and_cleanup_files(
inFiles, outFiles,
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem))
fasta = readfasta.readFasta(inputFasta, key="full", strip=False)
sequenceIDs = {}
for fastaID in fasta.keys():
shortID = fastaID.strip().strip('>').split(' ')[0]
sequenceIDs[shortID] = fastaID.strip().strip('>')
values = {}
for row in open(
os.path.join(rawOutput, "%s.output" %
Path(inputFasta).stem)).read().split('\n'):
if row.startswith('#'):
tokens = row.split(' ')
if tokens[1] not in values.keys():
values[tokens[1]] = [""] * 4
if "Number of predicted TMHs:" in row:
values[tokens[1]][0] = tokens[-1].strip()
if "Exp number of AAs in TMHs:" in row:
values[tokens[1]][1] = tokens[-1].strip()
if "Exp number, first 60 AAs:" in row:
values[tokens[1]][2] = tokens[-1].strip()
if "Total prob of N-in:" in row:
values[tokens[1]][3] = tokens[-1].strip()
output = [
"ID\tPredicted_TMH#\tExp_AAs#\tExp_first_60_AAs#\tTotal_N-in_prob"
]
for shortID in sequenceIDs.keys():
if shortID in values.keys():
output.append('\t'.join([sequenceIDs[shortID]] +
values[shortID]))
else:
output.append('\t'.join([sequenceIDs[shortID]] + [""] * 4))
open(os.path.join(rawOutput, "%s.tmhmm.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
if not rawFlag:
os.remove(
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem))
for tmpFile in glob.glob(os.path.join(os.getcwd(), "TMHMM_*")):
shutil.rmtree(tmpFile)
def run_signalp(inputFasta, outputDir, organism, multiFlag, process,
rawFlag):
command = SIGNALP_PATH
rawOutput = os.path.join(outputDir, "SIGNALP")
if not os.path.exists(rawOutput):
os.mkdir(rawOutput)
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
if multiFlag.lower() in ['t', 'true'
] and len(sequenceIDs) > SPLIT_LIMIT:
(inFiles, outFiles) = Feature.split_files(inputFasta, rawOutput)
else:
inFiles = [inputFasta]
outFiles = [
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem)
]
clines = []
for (f, inFile) in enumerate(inFiles):
outFile = outFiles[f]
if organism.lower() in ["gram+", "g+"]:
cline = "%s -t gram+ %s > %s" % (command, inFile, outFile)
elif organism.lower() in ["gram-", "g-"]:
cline = "%s -t gram- %s > %s" % (command, inFile, outFile)
clines.append(cline)
Feature.mp_run(clines, process)
if multiFlag.lower() in ['t', 'true'
] and len(sequenceIDs) > SPLIT_LIMIT:
Feature.combine_and_cleanup_files(
inFiles, outFiles,
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem))
fasta = readfasta.readFasta(inputFasta, key="full", strip=False)
sequenceIDs = {}
for fastaID in fasta.keys():
shortID = fastaID.strip().strip('>').split(' ')[0]
sequenceIDs[shortID] = fastaID.strip().strip('>')
values = {}
for (i, row) in enumerate(
open(
os.path.join(rawOutput, "%s.output" %
Path(inputFasta).stem)).readlines()):
if row == '' or row.startswith('#'):
continue
tokens = re.split('[ ]+', row)
if tokens[0] in values.keys():
print("Duplicate SingalP Result: %s" % tokens[0])
values[tokens[0]] = tokens[8]
output = ["ID\tSignalP_DScore"]
for shortID in sequenceIDs.keys():
if shortID in values.keys():
output.append('\t'.join(
[sequenceIDs[shortID], values[shortID]]))
else:
output.append('\t'.join([sequenceIDs[shortID], '0.000']))
open(os.path.join(rawOutput, "%s.signalp.tsv" % Path(inputFasta).stem),
'w').write('\n'.join(output))
if not rawFlag:
os.remove(
os.path.join(rawOutput, "%s.output" % Path(inputFasta).stem))
def train(self, positiveFasta, negativeFasta, outputDir, organism,
incFeatures, multiFlag, process):
featureDir = os.path.join(outputDir, "_FEATURE")
masterLabels = ["ID", "Label", "Gram"]
masterData = {}
for (g, inputFasta) in enumerate([negativeFasta, positiveFasta]):
sequenceIDs = readfasta.readFastaDesc(inputFasta, key="full")
if organism.lower() in ["gram+", "g+"]:
for fastaID in sequenceIDs:
masterData[fastaID] = [fastaID, str(g), "1"]
for method in incFeatures:
tsvFile = os.path.join(
featureDir, method.upper(),
"%s.%s.tsv" % (Path(inputFasta).stem, method))
for (i,
line) in enumerate(open(tsvFile).read().splitlines()):
tokens = line.split('\t')
if i == 0:
if method == "psortb":
masterLabels += tokens[2:]
else:
masterLabels += tokens[1:]
else:
if method == "psortb":
masterData[tokens[0]] += tokens[2:]
else:
masterData[tokens[0]] += tokens[1:]
elif organism.lower() in ["gram-", "g-"]:
for fastaID in sequenceIDs:
masterData[fastaID] = [fastaID, str(g), "0"]
for method in incFeatures:
tsvFile = os.path.join(
featureDir, method.upper(),
"%s.%s.tsv" % (Path(inputFasta).stem, method))
for (i,
line) in enumerate(open(tsvFile).read().splitlines()):
tokens = line.split('\t')
if i == 0:
if method == "psortb":
masterLabels += tokens[2:]
else:
masterLabels += tokens[1:]
else:
if method == "psortb":
masterData[tokens[0]] += tokens[2:]
else:
masterData[tokens[0]] += tokens[1:]
rows = []
samples = []
groups = []
for fastaID in masterData.keys():
tokens = masterData[fastaID]
samples.append(fastaID)
groups.append(int(tokens[1]))
rows.append(tokens[2:])
X = np.array(rows)
X = X.astype(float)
y = np.array(groups)
scaler = MinMaxScaler(copy=False)
scaler.fit(X)
joblib.dump(scaler, os.path.join(outputDir, "Scaler.sav"))
X = scaler.transform(X)
est = XGBClassifier(objective='binary:logistic',
silent=True,
nthread=1,
eval_metric='auc',
random_state=26)
estPipe = Pipeline([('feature_selection', SelectKBest(mRMR)),
('classification', est)])
grid = [{
"feature_selection__k": list(range(20, X.shape[1], 20))[:10],
'classification__learning_rate': [0.3, 0.1],
'classification__n_estimators': [60, 80, 100, 120, 140, 160],
'classification__max_depth': [3, 6, 9],
'classification__min_child_weight': [1, 3],
'classification__scale_pos_weight': [1, 6],
'classification__max_delta_step': [0, 3],
}]
cv = StratifiedShuffleSplit(n_splits=3, random_state=6)
if multiFlag.lower() not in ['t', 'true']:
xgb = GridSearchCV(estimator=estPipe,
param_grid=grid,
cv=cv,
iid=False,
verbose=1,
n_jobs=1)
else:
xgb = GridSearchCV(estimator=estPipe,
param_grid=grid,
cv=cv,
iid=False,
verbose=1,
n_jobs=process)
xgb.fit(X, y)
joblib.dump(xgb, os.path.join(outputDir, "VaxignML.sav"))
y_prob = xgb.predict_proba(X)
joblib.dump(y_prob, os.path.join(outputDir, "VaxignML.scores"))