def mostLikely(self, normal, island, dnastrand):
states = "NR"
alphabet = "AGTC"
normal = [float(x)/100 for x in normal]
island = [float(x)/100 for x in island]
p_initial = [1.0, 0.0]
p_initial = asarray(p_initial)
p_transition = []
p_transition.append([1.0-normal[-1], normal[-1]])
p_transition.append([island[-1], 1.0-island[-1]])
p_transition = asarray(p_transition)
p_emission = [] # 2x4 matrix
p_emission.append(normal[:4])
p_emission.append(island[:4])
p_emission = asarray(p_emission)
mm = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
x = MarkovModel.find_states(mm, dnastrand)
states, x = x[0]
return ''.join(states)
def mostLikely(self, normal, island, dnastrand):
states = "NR"
alphabet = "AGTC"
normal = [float(x)/100 for x in normal]
island = [float(x)/100 for x in island]
p_initial = [1.0, 0.0]
p_initial = asarray(p_initial)
p_transition = []
p_transition.append([1.0-normal[-1], normal[-1]])
p_transition.append([island[-1], 1.0-island[-1]])
p_transition = asarray(p_transition)
p_emission = [] # 2x4 matrix
p_emission.append(normal[:4])
p_emission.append(island[:4])
p_emission = asarray(p_emission)
mm = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
x = MarkovModel.find_states(mm, dnastrand)
states, x = x[0]
return ''.join(states)
def test_topcoder5(self):
# N
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.84, 0.16], [0.25, 0.75]])
p_emission = array([[0.26, 0.37, 0.08, 0.29], [0.31, 0.13, 0.33,
0.23]])
markov_model = MarkovModel.MarkovModel(states, alphabet, p_initial,
p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "T")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ["N"])
def test_topcoder1(self):
# NNNN
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.90, 0.10], [0.20, 0.80]])
p_emission = array([[0.30, 0.20, 0.30, 0.20], [0.10, 0.40, 0.10,
0.40]])
markov_model = MarkovModel.MarkovModel(states, alphabet, p_initial,
p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "TGCC")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ["N", "N", "N", "N"])
def test_topcoder5(self):
# N
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.84, 0.16],
[0.25, 0.75]])
p_emission = array([[0.26, 0.37, 0.08, 0.29],
[0.31, 0.13, 0.33, 0.23]])
markov_model = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "T")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ["N"])
def test_topcoder4(self):
# NRRRRRRRRRR
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.55, 0.45],
[0.15, 0.85]])
p_emission = array([[0.75, 0.03, 0.01, 0.21],
[0.34, 0.11, 0.39, 0.16]])
markov_model = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "TTAGCAGTGCG")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ['N', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R'])
def test_topcoder3(self):
# NRRRRRRRRRRRNNNNRRRRRRRRR
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.75, 0.25],
[0.25, 0.75]])
p_emission = array([[0.45, 0.36, 0.06, 0.13],
[0.24, 0.18, 0.12, 0.46]])
markov_model = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "CCGTACTTACCCAGGACCGCAGTCC")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ['N', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'N', 'N', 'N', 'N', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R', 'R'])
def test_topcoder2(self):
# NNNRRRNNRRNRRN
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.56, 0.44],
[0.25, 0.75]])
p_emission = array([[0.04, 0.14, 0.62, 0.20],
[0.39, 0.15, 0.04, 0.42]])
markov_model = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "CCTGAGTTAGTCGT")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ['N', 'N', 'N', 'R', 'R', 'R', 'N', 'N', 'R', 'R', 'N', 'R', 'R', 'N'])
def test_topcoder1(self):
# NNNN
states = "NR"
alphabet = "AGTC"
p_initial = array([1.0, 0.0])
p_transition = array([[0.90, 0.10],
[0.20, 0.80]])
p_emission = array([[0.30, 0.20, 0.30, 0.20],
[0.10, 0.40, 0.10, 0.40]])
markov_model = MarkovModel.MarkovModel(
states, alphabet, p_initial, p_transition, p_emission)
states = MarkovModel.find_states(markov_model, "TGCC")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ['N', 'N', 'N', 'N'])
def _eval_mm(self, motif_num=1, seq=''):
"""Return log_score_list of a sequence according to motif's HMM."""
mm = self.hmms_list[motif_num - 1]
hidden_states = len(mm.states)
seq_len = len(seq)
if seq_len < hidden_states:
raise ValueError('Sequence must be at least as long as the motif')
score = list()
for i in range(seq_len - hidden_states + 1):
seq_segment = seq[i:i + hidden_states - 1]
result = MarkovModel.find_states(mm, seq_segment)
score.append(result[0][1])
eps = 1e-100
log_score = [math.log(x + eps) for x in score]
# zero padding
for i in range(len(seq) - len(score)):
log_score.append(0)
return log_score
def _eval_mm(self, motif_num=1, seq=''):
"""Return log_score_list of a sequence according to motif's HMM."""
mm = self.hmms_list[motif_num - 1]
hidden_states = len(mm.states)
seq_len = len(seq)
if seq_len < hidden_states:
raise ValueError('Sequence must be at least as long as the motif')
score = list()
for i in range(seq_len - hidden_states + 1):
seq_segment = seq[i:i + hidden_states - 1]
result = MarkovModel.find_states(mm, seq_segment)
score.append(result[0][1])
eps = 1e-100
log_score = [math.log(x + eps) for x in score]
# zero padding
for i in range(len(seq) - len(score)):
log_score.append(0)
return log_score
def _get_occurence_indexandscore_mm(self, seq, motif_num):
mm_i = self.hmms_list[motif_num]
seq_len = len(seq)
motif_len = len(mm_i.states)
scores = list()
start_indexes = list()
for i in range(seq_len - motif_len + 1):
segment_score = 0
for j in range(motif_len):
letter = seq[i + j]
segment_score += MarkovModel.find_states(mm_i, letter)[0][1]
if segment_score > self.threshold:
scores.append(segment_score)
start_indexes.append(i + 1)
last_indexes = [i + motif_len for i in start_indexes]
data = zip(start_indexes, last_indexes, scores)
sorted_data = sorted(data, key=self._get_key, reverse=True)
top_result = sorted_data[:self.k]
return top_result
def _get_occurence_indexandscore_mm(self, seq, motif_num):
mm_i = self.hmms_list[motif_num]
seq_len = len(seq)
motif_len = len(mm_i.states)
scores = list()
start_indexes = list()
for i in range(seq_len - motif_len + 1):
segment_score = 0
for j in range(motif_len):
letter = seq[i + j]
segment_score += MarkovModel.find_states(mm_i, letter)[0][1]
if segment_score > self.threshold:
scores.append(segment_score)
start_indexes.append(i + 1)
last_indexes = [i + motif_len for i in start_indexes]
data = zip(start_indexes, last_indexes, scores)
sorted_data = sorted(data, key=self._get_key, reverse=True)
top_result = sorted_data[:self.k]
return top_result
def test_train_visible(self):
states = ["0", "1", "2", "3"]
alphabet = ["A", "C", "G", "T"]
training_data = [
("AACCCGGGTTTTTTT", "001112223333333"),
("ACCGTTTTTTT", "01123333333"),
("ACGGGTTTTTT", "01222333333"),
("ACCGTTTTTTTT", "011233333333"),
]
markov_model = MarkovModel.train_visible(states, alphabet, training_data)
states = MarkovModel.find_states(markov_model, "AACGTT")
self.assertEqual(len(states), 1)
state_list, state_float = states[0]
self.assertEqual(state_list, ['0', '0', '1', '2', '3', '3'])
self.assertAlmostEqual(state_float, 0.0082128906)
self.assertEqual(markov_model.states, ['0', '1', '2', '3'])
self.assertEqual(markov_model.alphabet, ['A', 'C', 'G', 'T'])
self.assertEqual(len(markov_model.p_initial), 4)
self.assertAlmostEqual(markov_model.p_initial[0], 1.0)
self.assertAlmostEqual(markov_model.p_initial[1], 0.0)
self.assertAlmostEqual(markov_model.p_initial[2], 0.0)
self.assertAlmostEqual(markov_model.p_initial[3], 0.0)
self.assertEqual(len(markov_model.p_transition), 4)
self.assertEqual(len(markov_model.p_transition[0]), 4)
self.assertEqual(len(markov_model.p_transition[1]), 4)
self.assertEqual(len(markov_model.p_transition[2]), 4)
self.assertEqual(len(markov_model.p_transition[3]), 4)
self.assertAlmostEqual(markov_model.p_transition[0][0], 0.2)
self.assertAlmostEqual(markov_model.p_transition[0][1], 0.8)
self.assertAlmostEqual(markov_model.p_transition[0][2], 0.0)
self.assertAlmostEqual(markov_model.p_transition[0][3], 0.0)
self.assertAlmostEqual(markov_model.p_transition[1][0], 0.0)
self.assertAlmostEqual(markov_model.p_transition[1][1], 0.5)
self.assertAlmostEqual(markov_model.p_transition[1][2], 0.5)
self.assertAlmostEqual(markov_model.p_transition[1][3], 0.0)
self.assertAlmostEqual(markov_model.p_transition[2][0], 0.0)
self.assertAlmostEqual(markov_model.p_transition[2][1], 0.0)
self.assertAlmostEqual(markov_model.p_transition[2][2], 0.5)
self.assertAlmostEqual(markov_model.p_transition[2][3], 0.5)
self.assertAlmostEqual(markov_model.p_transition[3][0], 0.0)
self.assertAlmostEqual(markov_model.p_transition[3][1], 0.0)
self.assertAlmostEqual(markov_model.p_transition[3][2], 0.0)
self.assertAlmostEqual(markov_model.p_transition[3][3], 1.0)
self.assertEqual(len(markov_model.p_emission), 4)
self.assertEqual(len(markov_model.p_emission[0]), 4)
self.assertEqual(len(markov_model.p_emission[1]), 4)
self.assertEqual(len(markov_model.p_emission[2]), 4)
self.assertEqual(len(markov_model.p_emission[3]), 4)
self.assertAlmostEqual(markov_model.p_emission[0][0], 0.666667,
places=4)
self.assertAlmostEqual(markov_model.p_emission[0][1], 0.111111,
places=4)
self.assertAlmostEqual(markov_model.p_emission[0][2], 0.111111,
places=4)
self.assertAlmostEqual(markov_model.p_emission[0][3], 0.111111,
places=4)
self.assertAlmostEqual(markov_model.p_emission[1][0], 0.083333,
places=4)
self.assertAlmostEqual(markov_model.p_emission[1][1], 0.750000,
places=4)
self.assertAlmostEqual(markov_model.p_emission[1][2], 0.083333,
places=4)
self.assertAlmostEqual(markov_model.p_emission[1][3], 0.083333,
places=4)
self.assertAlmostEqual(markov_model.p_emission[2][0], 0.083333,
places=4)
self.assertAlmostEqual(markov_model.p_emission[2][1], 0.083333,
places=4)
self.assertAlmostEqual(markov_model.p_emission[2][2], 0.750000,
places=4)
self.assertAlmostEqual(markov_model.p_emission[2][3], 0.083333,
places=4)
self.assertAlmostEqual(markov_model.p_emission[3][0], 0.031250,
places=4)
self.assertAlmostEqual(markov_model.p_emission[3][1], 0.031250,
places=4)
self.assertAlmostEqual(markov_model.p_emission[3][2], 0.031250,
places=4)
self.assertAlmostEqual(markov_model.p_emission[3][3], 0.906250,
places=4)
alphabet = ["A", "C", "G", "T"]
training_data = [
("AACCCGGGTTTTTTT", "001112223333333"),
("ACCGTTTTTTT", "01123333333"),
("ACGGGTTTTTT", "01222333333"),
("ACCGTTTTTTTT", "011233333333"),
]
print "Training HMM"
mm = MarkovModel.train_visible(states, alphabet, training_data)
print "Classifying"
#print MarkovModel.find_states(mm, "AACGTT")
#Don't just print this, as the float may have different
#precision on different platforms. This returns a list
#containing a tuple containing a list (fine), and a float.
states = MarkovModel.find_states(mm, "AACGTT")
for state_list, state_float in states:
print "State %s, %0.10f" % (repr(state_list), state_float)
print_mm(mm)
print "TESTING baum welch"
states = ["CP", "IP"]
alphabet = ["cola", "ice_t", "lem"]
outputs = [
(2, 1, 0)
]
print "Training HMM"
p_initial = [1.0, 0.0000001]
alphabet = ["A", "C", "G", "T"]
training_data = [
("AACCCGGGTTTTTTT", "001112223333333"),
("ACCGTTTTTTT", "01123333333"),
("ACGGGTTTTTT", "01222333333"),
("ACCGTTTTTTTT", "011233333333"),
]
print "Training HMM"
mm = MarkovModel.train_visible(states, alphabet, training_data)
print "Classifying"
#print MarkovModel.find_states(mm, "AACGTT")
#Don't just print this, as the float may have different
#precision on different platforms. This returns a list
#containing a tuple containing a list (fine), and a float.
states = MarkovModel.find_states(mm, "AACGTT")
for state_list, state_float in states :
print "State %s, %0.10f" % (repr(state_list), state_float)
print_mm(mm)
print "TESTING baum welch"
states = ["CP", "IP"]
alphabet = ["cola", "ice_t", "lem"]
outputs = [
(2, 1, 0)
]
print "Training HMM"
p_initial = [1.0, 0.0000001]
