def main():
# Options
parser = argparse.ArgumentParser()
parser.add_argument('fasta',
help='''Fasta file of sequences for truncation.''')
parser.add_argument(
'weights',
help=
'''File containing model weights. This specifies which model to use.'''
)
parser.add_argument(
'output',
help='''File where results will be written. By default, this script will
not run if the output file already exists. Use the -o option to overwrite an existing file.'''
)
parser.add_argument(
'-o',
help=
'''Use this option if you want to overwrite an existing output file.''',
action='store_true')
parser.add_argument(
'-e',
help=
'''Use this option to write just the Ensembl transcript ID instead of
the full header.''',
action='store_true')
args = parser.parse_args()
##########
## MAIN ##
##########
lookup = dict(zip(range(5), 'NATCG'))
if not args.o and os.path.exists(args.output):
raise Exception(args.output + ' already exists!')
if args.e:
transpat = re.compile('ENST\d*.\d*')
print "Reading input files..."
full_seqs = fasta.load_fasta(args.fasta, 0)
mut = []
for seq, name in full_seqs:
for i in xrange(len(seq)):
for b in xrange(1, 5):
if b != seq[i]:
mut_seq = seq[:i] + [b] + seq[i + 1:]
mut.append((mut_seq, str(i), lookup[b], name))
mRNN = model.build_model(args.weights)
lines = []
print "Evaluating sequences..."
seqs, pos, base, name = zip(*mut)
if args.e:
name = [transpat.search(n).group() for n in name]
scores = mRNN.batch_predict(seqs, True)
scores = map(str, scores)
lines = zip(name, pos, base, scores)
lines = ['\t'.join(line) for line in lines]
with open(args.output, 'w') as out:
out.write('\n'.join(lines))
def train(posFastaFile, negFastaFile, posValFasta, negValFasta, parameters):
print "Reading input files..."
positives = fasta.load_fasta(posFastaFile, parameters['min_length'])
negatives = fasta.load_fasta(negFastaFile, parameters['min_length'])
valpos = fasta.load_fasta(posValFasta, parameters['min_length'])
valneg = fasta.load_fasta(negValFasta, parameters['min_length'])
train = positives, negatives
val = valpos, valneg
print "Building new model..."
mRNN = model.build_model(parameters['weights'],
parameters['embedding_size'],
parameters['recurrent_gate_size'], 5,
parameters['dropout'])
print inspect.getmodule(mRNN.__class__)
print "Training model..."
mRNN = model.train_model(mRNN, train, val, parameters['epochs'],
parameters['output'], parameters['max_length'],
parameters['save_freq'],
parameters['early_stopping'])
return mRNN
def load_file(input_filename, min_peptide_length=9, max_peptide_length=31):
"""
Load mutatated peptides from FASTA, VCF, or MAF file.
For the latter two formats, expand their variants across all
annotated transcripts.
Parameters
--------
input_filename : str
min_peptide_length : int
max_peptide_length : int
Returns a dataframe with columns:
- chr : chomosome
- pos : position in the chromosome
- ref : reference DNA
- alt : alternate DNA
- info : gene name and entrez gene ID
- stable_id_transcript : Ensembl transcript ID
- SourceSequence : region of protein around mutation
- MutationStart : first amino acid modified
- MutationEnd : last mutated amino acid
- GeneMutationInfo : original genetic variant e.g. chr3 g.484899 C>T
- PeptideMutationInfo : annotation e.g. V600E
"""
if input_filename.endswith(".fasta") \
or input_filename.endswith(".fa"):
return load_fasta(input_filename)
vcf_df = load_variants(input_filename)
vcf_df = vcf_df.drop_duplicates()
return expand_transcripts(
vcf_df,
input_filename,
min_peptide_length = min_peptide_length,
max_peptide_length = max_peptide_length)
def load_file(input_filename, min_peptide_length=9, max_peptide_length=31):
"""
Load mutatated peptides from FASTA, VCF, or MAF file.
For the latter two formats, expand their variants across all
annotated transcripts.
Parameters
--------
input_filename : str
min_peptide_length : int
max_peptide_length : int
Returns a dataframe with columns:
- chr : chomosome
- pos : position in the chromosome
- ref : reference DNA
- alt : alternate DNA
- info : gene name and entrez gene ID
- stable_id_transcript : Ensembl transcript ID
- SourceSequence : region of protein around mutation
- MutationStart : first amino acid modified
- MutationEnd : last mutated amino acid
- GeneMutationInfo : original genetic variant e.g. chr3 g.484899 C>T
- PeptideMutationInfo : annotation e.g. V600E
"""
if input_filename.endswith(".fasta") \
or input_filename.endswith(".fa"):
return load_fasta(input_filename)
vcf_df = load_variants(input_filename)
vcf_df = vcf_df.drop_duplicates()
return expand_transcripts(vcf_df,
input_filename,
min_peptide_length=min_peptide_length,
max_peptide_length=max_peptide_length)
def main():
# Options
opts, files = getopt.getopt(sys.argv[1:], "hvo:w:E:b:e:r:d:t:p:f:", [
"help",
"output=",
"weights=",
"epochs=",
"batch_size=",
"embedding_size=",
"recurrent_gate_size=",
"dropout=",
"test=",
"min_length=",
"max_length=",
"file_label=",
])
if len(files) != 1:
usage()
fastaFile = files[0]
print "using fasta file: ", fastaFile
# Defaults:
parameters = {}
parameters['output'] = None
parameters['verbose'] = False
parameters['weights'] = None
parameters['batch_size'] = 16
parameters['embedding_size'] = 128
parameters['recurrent_gate_size'] = 256
parameters['dropout'] = 0.1
parameters['test'] = 0.1
parameters['min_length'] = 200
parameters['max_length'] = 1000
parameters['num_train'] = 10000
parameters['epochs'] = 50
parameters['save_freq'] = 3
parameters['file_label'] = ""
# loop over options:
for option, argument in opts:
if option == "-v":
parameters[verbose] = True
elif option in ("-h", "--help"):
usage()
elif option in ("-o", "--output"):
parameters['output'] = argument
elif option in ("-w", "--weights"):
parameters['weights'] = argument
elif option in ("-E", "--epochs"):
parameters['epochs'] = int(argument)
elif option in ("-b", "--batch_size"):
parameters['batch_size'] = int(argument)
elif option in ("-e", "--embedding_size"):
parameters['embedding_size'] = int(argument)
elif option in ("-d", "--dropout"):
parameters['dropout'] = float(argument)
elif option in ("-t", "--test"):
parameters['test'] = float(argument)
elif option in ("-l", "--min_length"):
parameters['min_length'] = int(argument)
elif option in ("-L", "--max_length"):
parameters['max_length'] = int(argument)
elif option in ("-n", "--num_train"):
parameters['num_train'] = int(argument)
elif option in ("-f", "--file_label"):
parameters['file_label'] = argument
else:
assert False, "unhandled option"
##########
## MAIN ##
##########
print "Reading input files..."
sequences = fasta.load_fasta(fastaFile, parameters['min_length'])
if not parameters['weights']:
print "No weights given with -w parameter.\n"
sys.exit()
modelFiles = parameters['weights'].split(',')
models = []
for modelFile in modelFiles:
print "Building model..."
mRNN = model.build_model(modelFile, parameters['embedding_size'],
parameters['recurrent_gate_size'], 5,
parameters['dropout'])
models.append(mRNN)
print "Evaluating sequences..."
output = fastaFile + ".mRNNensemble"
if parameters['output']:
output = parameters['output']
evaluate.ensemble_evaluate_sequences(models, sequences, output,
parameters['batch_size'])
def main():
# Options
opts, files = getopt.getopt(sys.argv[1:], "hvo:w:E:b:e:r:d:t:l:L:s:", [
"help", "output=", "weights=", "epochs=", "batch_size=",
"embedding_size=", "recurrent_gate_size=", "dropout=", "test=",
"min_length=", "max_length=", "early_stopping="
])
if len(files) != 4:
usage()
posFastaFile = files[0]
negFastaFile = files[1]
posValFasta = files[2]
negValFasta = files[3]
print "using positive file: ", posFastaFile
print "using negative file: ", negFastaFile
print "using positive validation file: ", posValFasta
print "using negative validation file: ", negValFasta
# Defaults:
parameters = {}
parameters['output'] = None
parameters['verbose'] = False
parameters['weights'] = None
parameters['batch_size'] = 16
parameters['embedding_size'] = 128
parameters['recurrent_gate_size'] = 256
parameters['dropout'] = 0.5
parameters['test'] = 0.1
parameters['min_length'] = 200
parameters['max_length'] = 1000
parameters['num_train'] = 10000
parameters['epochs'] = 25
parameters['save_freq'] = 1
parameters['early_stopping'] = None
# loop over options:
for option, argument in opts:
if option == "-v":
parameters[verbose] = True
elif option in ("-h", "--help"):
usage()
elif option in ("-o", "--output"):
parameters['output'] = argument
elif option in ("-w", "--weights"):
parameters['weights'] = argument
elif option in ("-E", "--epochs"):
parameters['epochs'] = int(argument)
elif option in ("-b", "--batch_size"):
parameters['batch_size'] = int(argument)
elif option in ("-e", "--embedding_size"):
parameters['embedding_size'] = int(argument)
elif option in ("-r", "--recurrent_gate_size"):
parameters['recurrent_gate_size'] = int(argument)
elif option in ("-d", "--dropout"):
parameters['dropout'] = float(argument)
elif option in ("-t", "--test"):
parameters['test'] = float(argument)
elif option in ("-l", "--min_length"):
parameters['min_length'] = int(argument)
elif option in ("-L", "--max_length"):
parameters['max_length'] = int(argument)
elif option in ("-n", "--num_train"):
parameters['num_train'] = int(argument)
elif option in ("-s", "--early_stopping"):
if argument is not None:
argument = int(argument)
parameters['early_stopping'] = argument
else:
assert False, "unhandled option"
##########
## MAIN ##
##########
print "Reading input files..."
positives = fasta.load_fasta(posFastaFile, parameters['min_length'])
negatives = fasta.load_fasta(negFastaFile, parameters['min_length'])
valpos = fasta.load_fasta(posValFasta, parameters['min_length'])
valneg = fasta.load_fasta(negValFasta, parameters['min_length'])
train = positives, negatives
val = valpos, valneg
print "Building new model..."
mRNN = model.build_model(parameters['weights'],
parameters['embedding_size'],
parameters['recurrent_gate_size'], 5,
parameters['dropout'])
print inspect.getmodule(mRNN.__class__)
print "Training model..."
mRNN = model.train_model(mRNN, train, val, parameters['epochs'],
parameters['output'], parameters['max_length'],
parameters['save_freq'],
parameters['early_stopping'])
return mRNN
def mutantsFromFasta(inputFile):
'Loads the first sequence from the given file, mutates it.'
seq = fasta.load_fasta(inputFile, 0)[0][0]
seqs = zip(*pairMutate(seq))
return seq, seqs
def load_test_seqs():
seqs = load_fasta("resources/test.fa")
return seqs
def main():
# Options
opts, files = getopt.getopt(sys.argv[1:], "hvo:w:E:b:e:r:d:t:p:f:", ["help",
"output=",
"weights=",
"epochs=",
"batch_size=",
"embedding_size=",
"recurrent_gate_size=",
"dropout=",
"test=",
"min_length=",
"max_length=",
"file_label=",
])
if len(files) != 2:
usage()
posFastaFile = files[0]
negFastaFile = files[1]
print "using positive file: ", posFastaFile
print "using negative file: ", negFastaFile
# Defaults:
parameters = {}
parameters['output'] = None
parameters['verbose'] = False
parameters['weights'] = None
parameters['batch_size'] = 16
parameters['embedding_size'] = 128
parameters['recurrent_gate_size'] = 256
parameters['dropout'] = 0.1
parameters['test'] = 0.1
parameters['min_length'] = 200
parameters['max_length'] = 1000
parameters['num_train'] = 10000
parameters['epochs'] = 50
parameters['save_freq'] = 3
parameters['file_label'] = ""
# loop over options:
for option, argument in opts:
if option == "-v":
parameters[verbose] = True
elif option in ("-h", "--help"):
usage()
elif option in ("-o", "--output"):
parameters['output'] = argument
elif option in ("-w", "--weights"):
parameters['weights'] = argument
elif option in ("-E", "--epochs"):
parameters['epochs'] = int(argument)
elif option in ("-b", "--batch_size"):
parameters['batch_size'] = int(argument)
elif option in ("-e", "--embedding_size"):
parameters['embedding_size'] = int(argument)
elif option in ("-d", "--dropout"):
parameters['dropout'] = float(argument)
elif option in ("-t", "--test"):
parameters['test'] = float(argument)
elif option in ("-l", "--min_length"):
parameters['min_length'] = int(argument)
elif option in ("-L", "--max_length"):
parameters['max_length'] = int(argument)
elif option in ("-n", "--num_train"):
parameters['num_train'] = int(argument)
elif option in ("-f", "--file_label"):
parameters['file_label'] = argument
else:
assert False, "unhandled option"
##########
## MAIN ##
##########
print "Reading input files..."
positives = fasta.load_fasta(posFastaFile,parameters['min_length'])
negatives = fasta.load_fasta(negFastaFile,parameters['min_length'])
test = positives,negatives
print "Building model..."
if not parameters['weights']:
print "No weights given with -w parameter.\n"
sys.exit()
mRNN = model.build_model(parameters['weights'],parameters['embedding_size'],parameters['recurrent_gate_size'],5,parameters['dropout'])
print "Evaluating model..."
conf_mat = evaluate.evaluate_model(mRNN, test, parameters['batch_size'])
acc = evaluate.process_results(conf_mat,parameters)
def main():
# Options
parser = argparse.ArgumentParser(
description=
'''Takes a fasta file and a weights file for an RNN model as input.
After loading the RNN, each transcript in the fasta file is truncated at every possible position and the model
predicts the score. The output is a tab-delimited file with the following fields: transcript name, truncation position,
the model's prediction that the truncated sequence is coding, the log odds of that probability, and information about
where the position is in the transcript (5' UTR, CDS, 3' UTR, or none). If the -s
option is used, the transcipt name is in the filename, so the field is eliminated.'''
)
parser.add_argument('fasta',
help='''Fasta file of sequences for truncation.''')
parser.add_argument(
'weights',
help=
'''File containing model weights. This specifies which model to use.'''
)
parser.add_argument(
'output',
help=
'''Output name. By default, this is the file where results will be written.
If using the -s option, it is the directory where results will be written. By default, this script will
not run if the output file or directory already exists. Use the -o option to overwrite an existing file,
or to potentially overwrite files in the output directory if using the -s option. Note that with the -s option,
file names are chosen based on the defline of the transcript. If the -s and -o options are used together,
files in the output directory will not be deleted, but may (or may not) be overwritten.'''
)
parser.add_argument(
'-o',
help=
'''Use this option if you want to overwrite an existing output file, or use an
existing output directory, potentially (but not certainly) overwriting files in it.''',
action='store_true')
parser.add_argument(
'-e',
help=
'''Use this option to write just the Ensembl transcript ID instead of
the full defline.''',
action='store_true')
parser.add_argument(
'-s',
help=
'''Use this option to split the output into individual files named based on the
defline of the transcript. If using this option, the output argument should be the name of a directory,
not a file.''',
action='store_true')
args = parser.parse_args()
##########
## MAIN ##
##########
if not args.o and os.path.exists(args.output):
if args.s:
field1 = 'directory'
else:
field1 = 'file'
raise Exception(
args.output +
''' already exists! Please choose a different {0} name or use the -o option to overwrite. Use the command python {1} -h for more details.'''
.format(field1, sys.argv[0]))
orig_dir = os.getcwd()
if args.e:
transpat = re.compile('ENST\d*.\d*')
cds_loc = re.compile('CDS:(\d+)-(\d+)')
print "Reading input files..."
full_seqs = fasta.load_fasta(args.fasta, 0)
if args.s:
try:
os.mkdir(args.output)
except OSError:
#os.mkdir raises this error if the path already exists. We should not see this error unless using the args.o option.
#if the assertion is true, then everything's ok, and we can just ignore the error and move to the directory since it already exists.
assert args.o == True
trunc = []
for seq, name in full_seqs:
coords = cds_loc.search(name)
if coords:
#there is a CDS field in the defline
coords = map(int, coords.group(1, 2))
for i in xrange(len(seq)):
if not coords:
pos_class = 'NA'
elif i < coords[
0] - 1: #subtract one because info in defline is 1-based
pos_class = 'UTR5'
elif i < coords[1]:
pos_class = 'CDS'
else:
pos_class = 'UTR3'
trunc.append((seq[:i + 1], str(i), name, pos_class))
mRNN = model.build_model(args.weights)
print "Evaluating sequences..."
seqs, pos, name, pos_class = zip(*trunc)
cds_coords = []
if args.e:
names = [transpat.search(n).group() for n in name]
else:
names = [line.strip() for line in name]
probs = mRNN.batch_predict(seqs)
logodds = logit(probs)
#stringify numbers
probs = map(str, probs)
logodds = map(str, logodds)
if args.s:
os.chdir(args.output)
lines = zip(pos, probs, logodds, pos_class)
lines = ['\t'.join(line) for line in lines]
#put lines in arrays keyed by transcript name
linedict = {name: [] for name in set(names)}
for name, line in zip(names, lines):
linedict[name].append(line)
for name in linedict:
with open(name + '.trunc.txt', 'w') as out:
out.write('\n'.join(linedict[name]))
#go back to original directory
os.chdir(orig_dir)
else:
lines = zip(names, pos, probs, logodds, pos_class)
lines = ['\t'.join(line) for line in lines]
with open(args.output, 'w') as out:
out.write('\n'.join(lines))
def main():
# Options
opts, files = getopt.getopt(sys.argv[1:], "hvo:w:E:b:e:r:d:t:p:f:", [
"help",
"output=",
"weights=",
"epochs=",
"batch_size=",
"embedding_size=",
"recurrent_gate_size=",
"dropout=",
"test=",
"min_length=",
"max_length=",
"file_label=",
])
if len(files) != 2:
usage()
posFastaFile = files[0]
negFastaFile = files[1]
print "using positive file: ", posFastaFile
print "using negative file: ", negFastaFile
# Defaults:
parameters = {}
parameters['output'] = None
parameters['verbose'] = False
parameters['weights'] = None
parameters['batch_size'] = 16
parameters['embedding_size'] = 128
parameters['recurrent_gate_size'] = 256
parameters['dropout'] = 0.1
parameters['test'] = 0.1
parameters['min_length'] = 200
parameters['max_length'] = 1000
parameters['num_train'] = 10000
parameters['epochs'] = 50
parameters['save_freq'] = 3
parameters['file_label'] = ""
# loop over options:
for option, argument in opts:
if option == "-v":
parameters[verbose] = True
elif option in ("-h", "--help"):
usage()
elif option in ("-o", "--output"):
parameters['output'] = argument
elif option in ("-w", "--weights"):
parameters['weights'] = argument
elif option in ("-E", "--epochs"):
parameters['epochs'] = int(argument)
elif option in ("-b", "--batch_size"):
parameters['batch_size'] = int(argument)
elif option in ("-e", "--embedding_size"):
parameters['embedding_size'] = int(argument)
elif option in ("-d", "--dropout"):
parameters['dropout'] = float(argument)
elif option in ("-t", "--test"):
parameters['test'] = float(argument)
elif option in ("-l", "--min_length"):
parameters['min_length'] = int(argument)
elif option in ("-L", "--max_length"):
parameters['max_length'] = int(argument)
elif option in ("-n", "--num_train"):
parameters['num_train'] = int(argument)
elif option in ("-f", "--file_label"):
parameters['file_label'] = argument
else:
assert False, "unhandled option"
##########
## MAIN ##
##########
print "Reading input files..."
positives = fasta.load_fasta(posFastaFile, parameters['min_length'])
negatives = fasta.load_fasta(negFastaFile, parameters['min_length'])
test = positives, negatives
if not parameters['weights']:
print "No weights given with -w parameter.\n"
sys.exit()
modelFiles = parameters['weights'].split(',')
models = []
for modelFile in modelFiles:
print "Building model..."
mRNN = model.build_model(modelFile, parameters['embedding_size'],
parameters['recurrent_gate_size'], 5,
parameters['dropout'])
models.append(mRNN)
print "Evaluating model..."
conf_mat = evaluate.evaluate_multi_model(models, test,
parameters['batch_size'])
[[TN, FP], [FN, TP]] = conf_mat
acc = (TP + TN) / (TP + TN + FP + FN)
sens = TP / (TP + FN)
spec = TN / (TN + FP)
outFile = "multi_predict" + ".acc.txt"
if parameters['file_label']:
outFile = "multi_predict" + "." + parameters['file_label'] + ".acc.txt"
F = open(outFile, 'w')
F.write("%s\tACC\t%.4f\n" % (parameters['weights'], acc))
F.write("%s\tSPEC\t%.4f\n" % (parameters['weights'], spec))
F.write("%s\tSENS\t%.4f\n" % (parameters['weights'], sens))
F.write("%d\t%d\n%d\t%d\n" % (TN, FP, FN, TP))
F.close()
def main():
# Options
parser = argparse.ArgumentParser(
description=
'''Takes a fasta file and a weights file for an RNN model as input.
After loading the RNN, the 3' UTR, CDS, and 5' UTR are individually shuffled, and the entire transcript is scored.
The number of shuffles done is determined with the -n option. The model predicts the probability that the shuffled
and unshuffled transcripts are protein coding, and Z-scores are computed for the 3' UTR, CDS, and 5' UTR.
The output is a tab-delimited file with the following fields: transcript name, 3' UTR Z-score, CDS Z-score, 5' UTR Z-score,
3' UTR length, CDS length, 5' UTR length.''')
parser.add_argument('fasta',
help='''Fasta file of sequences for shuffling.''')
parser.add_argument(
'weights',
help=
'''File containing model weights. This specifies which model to use.'''
)
parser.add_argument(
'output',
help='''Output name. This is the file where results will be written.
By default, this script will not run if the output file already exists. Use the -o option to overwrite an existing file.'''
)
parser.add_argument(
'-o',
help=
'''Use this option if you want to overwrite an existing output file.''',
action='store_true')
parser.add_argument(
'-e',
help=
'''Use this option to write just the Ensembl transcript ID instead of
the full defline.''',
action='store_true')
parser.add_argument(
'-n',
help='''Number of times to shuffle each segment. Default 20.''',
default=20,
type=int)
parser.add_argument(
'-p',
help=
'''Base of filenames to plot to. The plot is a matplotlib plot of z-score
vs sequence length. The -o option also applies to this file.''')
args = parser.parse_args()
##########
## MAIN ##
##########
if not args.o:
if os.path.exists(args.output):
field1 = 'file'
raise Exception(
args.output +
''' already exists! Please choose a different {0} name or use the -o option to overwrite. Use the command python {1} -h for more details.'''
.format(field1, sys.argv[0]))
if args.p and os.path.exists(args.p):
raise Exception(
args.p +
''' already exists! Please choose a different {0} name or use the -o option to overwrite. Use the command python {1} -h for more details.'''
.format(field1, sys.argv[0]))
orig_dir = os.getcwd()
if args.e:
transpat = re.compile('ENST\d*.\d*')
#never mind, CDS is sufficient
cds_loc = re.compile('CDS:(\d+)-(\d+)')
print "Reading input files..."
full_seqs = fasta.load_fasta(args.fasta, 0)
seqs = []
seq_lens = []
for seq, name in full_seqs:
coords = cds_loc.search(name)
#do shuffling here
if coords:
#there is a CDS field in the defline
coords = map(int, coords.group(1, 2))
else:
print name
continue
utr5 = seq[:coords[0] - 1]
cds = seq[coords[0] - 1:coords[1]]
utr3 = seq[coords[1]:]
seqs.append((seq, 'orig', name))
utr5_shuffle = [
shuf_utr5 + cds + utr3 for shuf_utr5 in shuffle_seq(utr5, args.n)
]
cds_shuffle = [
utr5 + shuf_cds + utr3 for shuf_cds in shuffle_seq(cds, args.n)
]
utr3_shuffle = [
utr5 + cds + shuf_utr3 for shuf_utr3 in shuffle_seq(utr3, args.n)
]
for seq_type, group in zip(('utr5', 'cds', 'utr3'),
(utr5_shuffle, cds_shuffle, utr3_shuffle)):
for s in group:
seqs.append((s, seq_type, name))
seq_lens.append(map(len, [utr5, cds, utr3]))
mRNN = model.build_model(args.weights)
print "Evaluating sequences..."
seqs, seq_type, name = zip(*seqs)
cds_coords = []
if args.e:
names = [transpat.search(n).group() for n in name]
else:
names = [line.strip() for line in name]
probs = mRNN.batch_predict(seqs)
#delete sequences here, since they are not needed anymore
del seqs
logodds = logit(probs)
Zscores = []
#calculate z-scores
i = 0
while i < len(probs):
batch = []
curr_name = names[i]
while curr_name == names[i]:
batch.append((probs[i], seq_type[i]))
i += 1
if i == len(probs):
break
assert batch[0][1] == 'orig'
orig = batch[0][0]
tmp = {'name': curr_name, 'utr5': None, 'cds': None, 'utr3': None}
for j in xrange(3):
try:
sub_batch = batch[1 + j * args.n:1 + (j + 1) * args.n]
tmp[sub_batch[0][1]] = z_score(orig, zip(*sub_batch)[0])
except IndexError:
pass
Zscores.append([tmp[key] for key in ['name', 'utr5', 'cds', 'utr3']])
comments = '#fasta: {0}, weights: {1}, number of shuffles: {2}'.format(
args.fasta, args.weights, args.n)
lines = [
comments,
"transcript\t5' UTR Z-score\tCDS Z-score\t3' UTR Z-score\t5' UTR length\tCDS length\t3' UTR length"
]
for z, lens in zip(Zscores, seq_lens):
line = map(str, z + lens)
lines.append('\t'.join(line))
with open(args.output, 'w') as out:
out.write('\n'.join(lines))
if args.p:
plot_zscore_scatter(Zscores, seq_lens, args.p)
plot_zscore_histogram(Zscores, seq_lens, args.p)