def hmm_iterator(hmms, sequences):
"""
Given a list of HMMs and a set of sequences, print the sensitivity, specifity and
accuracy of the models.
"""
for hmm in hmms:
#Set values to zero
sensitivity = 0
specificity = 0
accuracy = 0
hmm_name = hmm.split("/")[-1].split(".")[0]
#Parse tgf and run viterbi algorithm
hmm = tgf.parse(hmm)
alignments = viterbi.viterbi_all(hmm, sequences)
#Compute true and false positives and false negatives
true_positives = len(
filter(lambda align: align.state_path[50] == "*", alignments))
false_negatives = len(
filter(lambda align: align.state_path[50] != "*", alignments))
false_positives = false_negatives # True only for this concrete project!
#Calculate sensitivity, specificity and accuracy
sensitivity = float(true_positives) / (true_positives +
false_negatives)
specificity = float(true_positives) / (true_positives +
false_positives)
accuracy = (specificity + sensitivity) / 2.0
def print_percent(number):
number = "{:.4f} %".format(number * 100.0)
return number
print "{}\tSN: {}\tSP: {}\tavgSNSP: {}".format(
hmm_name, print_percent(sensitivity), print_percent(specificity),
print_percent(accuracy))
def subject(self, filename="simple.tgf"):
return tgf.parse("test/files/tgf/{}".format(filename))
else:
if is_correct:
fn += 1
else:
tn += 1
return (tp, tn, fp, fn)
def print_roc_data_in_tsv(roc_data):
print "score\ttpr\tfpr\tppv\ttp\ttn\tfp\tfn"
for score, metrics in sorted(roc_data.items()):
print "{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}\t{:.4f}".format(
score, *metrics)
hmm = tgf_parser.parse(sys.argv[1])
sequences = seq_parser.parse(sys.argv[2])
alignments = viterbi.viterbi_all(hmm, sequences)
evaluated_alignments = map(evaluate_alignment, alignments)
max_score = max(alignments, key=lambda align: align.score).score
min_score = min(alignments, key=lambda align: align.score).score
roc_data = {}
step_size = (max_score - min_score) / STEPS
scores_iterator = frange(float_floor(min_score), float_ceil(max_score),
step_size)
def subject(self, hmm=None, observations=10):
if hmm is None: hmm = tgf.parse("test/files/tgf/simple.tgf")
return sample.sample(hmm, observations)
return (action, args)
def get_output_filename_from_args(args, position):
if len(args) > position:
return args[position]
if __name__ == '__main__':
action, args = parse_args()
if action == 'sample':
if len(args) < 2: print_help_message(invalid=True)
hmm = tgf.parse(args[1])
length_of_sample = int(args[0])
output_filename = get_output_filename_from_args(args, 2)
mallet_writer.write([sample.sample(hmm, length_of_sample)],
output_filename)
elif action == 'viterbi':
if len(args) < 2: print_help_message(invalid=True)
hmm = tgf.parse(args[0])
sequences = seq_parser.parse(args[1])
output_filename = get_output_filename_from_args(args, 2)
mallet_writer.write(viterbi.viterbi_all(hmm, sequences),
output_filename)
def subject(self, hmm_filename, sequence):
return viterbi.viterbi(
tgf.parse("test/files/tgf/{}".format(hmm_filename)), sequence)