def predict(sequence, model_or_filelike, compute_posterior=True):
if isinstance(model_or_filelike, tuple):
model = model_or_filelike
else:
_, model = parse(open(model_or_filelike))
_, path = viterbi(sequence, *model)
if compute_posterior:
forward_table, constants = forward(sequence, *model)
backward_table = backward(sequence, constants, *model)
posterior = forward_table * backward_table
_, _, _, char_map, label_map, name_map = model
observations = len(sequence)
states = len(name_map)
# just counts how many states there are per label
group_counts = Counter(label_map.values())
table = np.zeros(shape=(observations, 3))
for i in range(observations):
group_probs = defaultdict(float)
for j in range(states):
group = label_map[j].lower()
group_probs[group] += posterior[i, j]
for k, group in enumerate(GROUP_NAMES):
table[i, k] = group_probs[group]
return path, table / table.sum(axis=1, keepdims=True)
return path
def predict(sequence, header, model_or_filelike, compute_posterior=True):
if isinstance(model_or_filelike, tuple):
model = model_or_filelike
else:
_, model = parse(open(model_or_filelike))
_, path = viterbi(sequence, *model)
if compute_posterior:
forward_table, constants = forward(sequence, *model)
backward_table = backward(sequence, constants, *model)
posterior = forward_table * backward_table
_, _, _, char_map, label_map, name_map = model
observations = len(sequence)
states = len(name_map)
table = np.zeros(shape=(observations, 3))
for i in range(observations):
group_probs = defaultdict(float)
for j in range(states):
group = label_map[j].lower()
group_probs[group] += posterior[i, j]
for k, group in enumerate(GROUP_NAMES):
table[i, k] = group_probs[group]
return path, table/table.sum(axis=1, keepdims=True)
return path
def cli():
parser = argparse.ArgumentParser()
parser.add_argument('-f', '--file', dest='sequence_file',
type=argparse.FileType('r'), required=True,
help='path to file in fasta format with sequences')
parser.add_argument('-m', '--model', dest='model_file',
type=argparse.FileType('r'), default=DEFAULT_MODEL,
help='path to the model to use')
if has_matplotlib:
parser.add_argument('-p', '--plot', dest='plot_posterior', action='store_true',
help='plot posterior probabilies')
args = parser.parse_args()
header, model = parse(args.model_file)
for entry in load_fasta_file(args.sequence_file):
path, posterior = predict(entry.sequence, header, model)
with open(entry.id + '.summary', 'w') as summary_file:
for start, end, state in summarize(path):
print("{} {} {}".format(start, end, PRETTY_NAMES[state]),
file=summary_file)
with open(entry.id + '.annotation', 'w') as ann_file:
print('>', entry.id, ' ', entry.description, sep='', file=ann_file)
for line in textwrap.wrap(path, 79):
print(line, file=ann_file)
plot_filename = entry.id + '.plot'
with open(plot_filename, 'w') as plot_file:
dump_posterior_file(plot_file, posterior)
if hasattr(args, 'plot_posterior') and args.plot_posterior:
with open(plot_filename, 'r') as fileobj:
plot(fileobj, entry.id + '.pdf')
def cli():
parser = argparse.ArgumentParser()
parser.add_argument('-f',
'--file',
dest='sequence_file',
type=argparse.FileType('r'),
required=True,
help='path to file in fasta format with sequences')
parser.add_argument('-m',
'--model',
dest='model_file',
type=argparse.FileType('r'),
default=DEFAULT_MODEL,
help='path to the model to use')
parser.add_argument('-p',
'--plot',
dest='plot_posterior',
action='store_true',
help='plot posterior probabilies')
args = parser.parse_args()
header, model = parse(args.model_file)
for record in sk.io.read(args.sequence_file, format='fasta'):
path, posterior = predict(normalized_sequence, header, model)
with open(record.metadata['id'] + '.summary', 'w') as summary_file:
for start, end, state in summarize(path):
print("{} {} {}".format(start, end, PRETTY_NAMES[state]),
file=summary_file)
with open(record.metadata['id'] + '.annotation', 'w') as ann_file:
print('>',
record.metadata['id'],
' ',
record.metadata['description'],
sep='',
file=ann_file)
for line in textwrap.wrap(path, 79):
print(line, file=ann_file)
plot_filename = record.metadata['id'] + '.plot'
with open(plot_filename, 'w') as plot_file:
print('inside', 'membrane', 'outside', file=plot_file)
for i in range(len(normalized_sequence)):
print('{} {} {}'.format(posterior[i, 0], posterior[i, 1],
posterior[i, 2]),
file=plot_file)
if args.plot_posterior:
plot(plot_filename, record.metadata['id'] + '.pdf')
def cli():
parser = argparse.ArgumentParser()
parser.add_argument('-f',
'--file',
dest='sequence_file',
type=argparse.FileType('r'),
required=True,
help='path to file in fasta format with sequences')
parser.add_argument('-m',
'--model',
dest='model_file',
type=argparse.FileType('r'),
default=DEFAULT_MODEL,
help='path to the model to use')
if has_matplotlib:
parser.add_argument('-p',
'--plot',
dest='plot_posterior',
action='store_true',
help='plot posterior probabilies')
args = parser.parse_args()
header, model = parse(args.model_file)
for entry in load_fasta_file(args.sequence_file):
path, posterior = predict(entry.sequence, header, model)
with open(entry.id + '.summary', 'w') as summary_file:
for start, end, state in summarize(path):
print("{} {} {}".format(start, end, PRETTY_NAMES[state]),
file=summary_file)
with open(entry.id + '.annotation', 'w') as ann_file:
print('>', entry.id, ' ', entry.description, sep='', file=ann_file)
for line in textwrap.wrap(path, 79):
print(line, file=ann_file)
plot_filename = entry.id + '.plot'
with open(plot_filename, 'w') as plot_file:
dump_posterior_file(plot_file, posterior)
if hasattr(args, 'plot_posterior') and args.plot_posterior:
with open(plot_filename, 'r') as fileobj:
plot(fileobj, entry.id + '.pdf')