def LDA(line_list, temp):
"""
:param line_list: list of SAM object
:param temp: temperature
:return:
"""
temp_list = [32, 37, 42, 47, 52, 57]
coef_list = [[[-0.14494789, 0.18791679, 0.02588474]],
[[-0.13364364, 0.22510179, 0.05494031]],
[[-0.09006122, 0.25660706, 0.1078303]],
[[-0.01593182, 0.24498485, 0.15753649]],
[[0.01860365, 0.1750174, 0.17003374]],
[[0.03236755, 0.11624593, 0.24306498]]]
inter_list = [-1.17545204, -5.40436344, -12.45549846,
-19.32670233, -20.11992898, -23.98652919]
class_list = [-1, 1]
try:
classfier_index = temp_list.index(temp)
except ValueError:
print("The given temperature was not in temp_list:", temp_list)
sys.exit()
coef_array = np.asarray(coef_list)
inter_array = np.asarray(inter_list)
class_array = np.asarray(class_list)
lda_classifer = LinearDiscriminantAnalysis()
lda_classifer.coef_ = coef_array[classfier_index]
lda_classifer.intercept_ = inter_array[classfier_index]
lda_classifer.classes_ = class_array
test_list = []
for sub_line in line_list:
if sub_line.xs_tag:
test_list.append([np.float(len(sub_line)), sub_line.xs_tag, sub_line.gc_content])
else:
return False
lda_prob = lda_classifer.predict_proba(np.asarray(test_list))[:, 1]
lda_prob = map(lambda x: x < 0.5, lda_prob)
if all(lda_prob):
return True
return False
# Convert lists to ndarrays. coefArray = np.asarray(coefList) interArray = np.asarray(interList) classArray = np.asarray(classList) # Determine which index to reference for model values. # default tempVal = 57 np_index = tempList.index(tempVal) # Build model from encoded values. clf = LinearDiscriminantAnalysis() clf.coef_ = coefArray[np_index] clf.intercept_ = interArray[np_index] clf.classes_ = classArray # Determine which classifier parameters to use. clfT = tempList.index(tempVal) # Make lists to hold data about candidates. testList = [] testSet = set() candsInfo = [] # the realignment result samf = "/Users/yeweijian/Downloads/data/t.sam" # Make a list to hold the output. outList = []
def cleanOutput(inputFile, uniqueVal, zeroVal, probVal, tempVal, sal, form,
reportVal, debugVal, metaVal, outNameVal, startTime):
# Determine the stem of the input filename.
fileName = str(inputFile).split('.')[0]
# Open input file for reading.
with open(inputFile, 'r') as f:
file_read = [line.strip() for line in f]
# Determine how many unique candidates are in the .sam file
samIDs = [x.split('\t')[0].split(':')[1].split('-')[0] \
if x[0] is not '@' else ' ' for x in file_read]
candsSet = set()
for x in samIDs:
if x is not ' ':
candsSet.add(x)
# Make a list to hold the output.
outList = []
# Make lists to hold Report info if desired.
if reportVal or debugVal is True:
rejectList = []
reportList = []
if uniqueVal or zeroVal is True:
# Process .sam file, keeping probes with only 0 or 1 unique alignment.
for i in range(0, len(file_read), 1):
if file_read[i][0] is not '@':
chromField = file_read[i].split('\t')[2]
chrom = file_read[i].split('\t')[0].split(':')[0]
start = file_read[i].split('\t')[0].split(':')[1].split('-')[0]
stop = file_read[i].split('\t')[0].split('-')[1].strip(' ')
seq = file_read[i].split('\t')[9]
Tm = probeTm(seq, sal, form)
# For unique mode.
if uniqueVal is True:
if re.match('\*', chromField) is None \
and re.search('XS', file_read[i]) is None:
outList.append('%s\t%s\t%s\t%s\t%s' \
% (chrom, start, stop, seq, Tm))
# Report info on selected probe if desired.
if reportVal is True:
reportList.append('Candidate probe at %s:%s-%s '
'aligned 1 time, added to output' \
% (chrom, start, stop))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s aligned 1 time, '
'added to output' % (chrom, start, stop))
else:
# Report info on rejected candidates if desired.
if reportVal or debugVal is True:
if start not in rejectList:
rejectList.append(start)
if re.match('\*', chromField) is not None:
if reportVal is True:
reportList.append('Candidate probe at '
'%s:%s-%s aligned 0 '
'times, was not added '
'to output' \
% (chrom, start, stop))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s '
'aligned 0 times, was not added to '
'output' % (chrom, start, stop))
elif re.search('XS', file_read[i]) is not None:
if reportVal is True:
reportList.append('Candidate probe at '
'%s:%s-%s aligned >1 '
'time, was not added '
'to output' \
% (chrom, start, stop))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s '
'aligned >1 time, was not added to '
'output' % (chrom, start, stop))
# For zero mode.
elif zeroVal is True:
if re.match('\*', chromField) is not None:
outList.append('%s\t%s\t%s\t%s\t%s' \
% (chrom, start, stop, seq, Tm))
# Report info on selected probe if desired.
if reportVal is True:
reportList.append('Candidate probe at %s:%s-%s '
'aligned 0 times, added to output '
'(Zero mode active)' \
% (chrom, start, stop))
if debugVal is True:
print('Candidate probe at %s:%s-%s aligned 0 times, '
'added to output (Zero mode active)' \
% (chrom, start, stop))
else:
# Report info on rejected candidates if desired.
if reportVal or debugVal is True:
if start not in rejectList:
rejectList.append(start)
if reportVal is True:
reportList.append('Candidate probe at '
'%s:%s-%s aligned >0 '
'times, was not added to '
'output (Zero mode '
'active)' \
% (chrom, start, stop))
if debugVal is True:
print('Candidate probe at %s:%s-%s aligned '
'>0 times, was not added to output '
'(Zero mode active)' \
% (chrom, start, stop))
# Else use LDA model.
else:
# Import scikit-learn LDA module.
# Note the module name changed between sklearn versions 0.16 and 0.17
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
# Import numpy module.
import numpy as np
# LDA model information.
tempList = [32, 37, 42, 47, 52, 57]
coefList = [[[-0.14494789, 0.18791679, 0.02588474]],
[[-0.13364364, 0.22510179, 0.05494031]],
[[-0.09006122, 0.25660706, 0.1078303]],
[[-0.01593182, 0.24498485, 0.15753649]],
[[0.01860365, 0.1750174, 0.17003374]],
[[0.03236755, 0.11624593, 0.24306498]]]
interList = [
-1.17545204, -5.40436344, -12.45549846, -19.32670233, -20.11992898,
-23.98652919
]
classList = [-1, 1]
# Convert lists to ndarrays.
coefArray = np.asarray(coefList)
interArray = np.asarray(interList)
classArray = np.asarray(classList)
# Determine which index to reference for model values.
np_index = tempList.index(tempVal)
# Build model from encoded values.
clf = LinearDiscriminantAnalysis()
clf.coef_ = coefArray[np_index]
clf.intercept_ = interArray[np_index]
clf.classes_ = classArray
# Determine which classifier parameters to use.
clfT = tempList.index(tempVal)
# Make lists to hold data about candidates.
testList = []
testSet = set()
candsInfo = []
# Process .sam file and extract information about each candidate probe.
for i in range(0, len(file_read), 1):
if file_read[i][0] is not '@':
chromField = file_read[i].split('\t')[2]
chrom = file_read[i].split('\t')[0].split(':')[0]
start = file_read[i].split('\t')[0].split(':')[1].split('-')[0]
stop = file_read[i].split('\t')[0].split('-')[1].strip(' ')
seq = file_read[i].split('\t')[9]
Tm = probeTm(seq, sal, form)
# First look for candidate probes with only one unique alignment.
if re.match('\*', chromField) is None \
and re.search('XS', file_read[i]) is None:
outList.append('%s\t%s\t%s\t%s\t%s' \
% (chrom, start, stop, seq, Tm))
# Record info on selected probe if desired.
if reportVal is True:
reportList.append('Candidate probe at %s:%s-%s aligned '
'1 time, added to output' \
% (chrom, start, stop))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s aligned 1 time, '
'added to output' % (chrom, start, stop))
# Populate lists that will be used to make the classification
# model input.
else:
if re.match('\*', chromField) is None \
and start not in testSet:
t = [
float(len(seq)),
float(file_read[i].split('\t')[12].split(':')[2]),
GC(seq)
]
testList.append(t)
testSet.add(start)
candsInfo.append('%s\t%s\t%s\t%s\t%s' \
% (chrom, start, stop, seq, Tm))
else:
# Report info on rejected candidates if desired.
if reportVal or debugVal is True:
if re.match('\*', chromField) is not None:
if start not in rejectList:
rejectList.append(start)
if reportVal is True:
reportList.append('Candidate probe at '
'%s:%s-%s aligned 0 '
'times, was not added '
'to output' \
% (chrom, start, stop))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s '
'aligned 0 times, was not added to '
'output' % (chrom, start, stop))
# Make ndarray for input into classifier.
testArray = np.asarray(testList)
# Create classifier
clf = LinearDiscriminantAnalysis()
# Load temperature-specific model information.
clf.coef_ = coefArray[clfT]
clf.intercept_ = interArray[clfT]
clf.classes_ = classArray
# Use model to predict the probability that candidate
# probes will have thermodynamically relevant
# off-target binding sites unless all have just 1
# alignment in the .sam file.
if len(testArray) > 1:
probs = clf.predict_proba(testArray)[:, 1]
# Filter through tested candidates using
# based on user-specified probability threshold.
for i in range(0, len(probs), 1):
if float(probs[i]) < probVal:
outList.append(candsInfo[i])
if reportVal is True:
reportList.append('Candidate probe at %s:%s-%s added to '
'output with %0.4f < %0.4f probability of '
'having off-target sites' \
% (candsInfo[i].split('\t')[0],
candsInfo[i].split('\t')[1],
candsInfo[i].split('\t')[2],
probs[i], probVal))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s added to output with '
'%0.4f < %0.4f probability of having off-target sites'
% (candsInfo[i].split('\t')[0],
candsInfo[i].split('\t')[1],
candsInfo[i].split('\t')[2], probs[i], probVal))
else:
if reportVal is True:
reportList.append('Candidate probe at %s:%s-%s filtered with '
'%0.4f => %0.4f probability of having '
'off-target sites' \
% (candsInfo[i].split('\t')[0],
candsInfo[i].split('\t')[1],
candsInfo[i].split('\t')[2],
probs[i], probVal))
if debugVal is True:
print(
'Candidate probe at %s:%s-%s filtered with '
'%0.4f => %0.4f probability of having off-target sites'
% (candsInfo[i].split('\t')[0],
candsInfo[i].split('\t')[1],
candsInfo[i].split('\t')[2], probs[i], probVal))
# Sort output list.
outList.sort(key=lambda x: [int(x.split('\t')[1])])
# Determine the name of the output file.
if outNameVal is None:
outName = '%s_probes' % fileName
else:
outName = outNameVal
# Create the output file.
output = open('%s.bed' % outName, 'w')
# Write the output file.
output.write('\n'.join(outList))
output.close()
# Print info about the results to terminal.
candsNum = len(candsSet)
cleanNum = len(outList)
if zeroVal is True:
print('outputClean identified %d of %d / %0.4f%% candidate probes as '
'having zero alignments' \
% (cleanNum, candsNum, float(cleanNum) / float(candsNum) * 100))
elif uniqueVal is True:
print(
'outputClean identified %d of %d / %0.4f%% candidate probes as '
'unique' %
(cleanNum, candsNum, float(cleanNum) / float(candsNum) * 100))
else:
print('outputClean passed %d of %d / %0.4f%% candidate probes through '
'specificity filtering using the %dC LDA model' \
% (cleanNum, candsNum,
float(cleanNum) / float(candsNum) * 100, tempVal))
# Write meta information to a .txt file if desired.
if metaVal is True:
metaText = open('%s_outputClean_meta.txt' % outName, 'w')
metaText.write('%s\t%f\t%s\t%d\t%d' \
% (inputFile,
timeit.default_timer() - startTime,
Version, cleanNum, candsNum))
metaText.close()
# If desired, create report file.
if reportVal is True:
reportOut = open('%s_outputClean_log.txt' % outName, 'w')
reportList.sort(key=lambda x: [int(x.split(':')[1].split('-')[0])])
reportList.insert(0,
'Results produced by %s %s' % (scriptName, Version))
reportList.insert(1, '-' * 100)
if uniqueVal is True:
reportList.insert(2, 'outputClean returned %d of %d / %0.4f%% '
'candidate probes as having exactly 1 '
'alignment' \
% (cleanNum, candsNum,
float(cleanNum) / float(candsNum) * 100))
elif zeroVal is True:
reportList.insert(2, 'outputClean returned %d of %d / %0.4f%% '
'candidate probes as having 0 alignments (Zero '
'mode active)' \
% (cleanNum, candsNum,
float(cleanNum) / float(candsNum) * 100))
else:
reportList.insert(2, 'outputClean passed %d of %d / %0.4f%% '
'candidate probes through specificity filtering '
'using the %dC LDA model' \
% (cleanNum, candsNum,
float(cleanNum) / float(candsNum) * 100,
tempVal))
reportList.insert(3, '-' * 100)
reportOut.write('\n'.join(reportList))
reportOut.close()
