1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
|
#!/usr/bin/env python
import StringIO
from operator import truth
from Bio import MarkovModel
from Numeric import ones, zeros, log, asarray
def print_mm(markov_model):
print "STATES: %s" % ' '.join(markov_model.states)
print "ALPHABET: %s" % ' '.join(markov_model.alphabet)
print "INITIAL:"
for i in range(len(markov_model.p_initial)):
print " %s: %.2f" % (
markov_model.states[i], markov_model.p_initial[i])
print "TRANSITION:"
for i in range(len(markov_model.p_transition)):
x = ["%.2f" % x for x in markov_model.p_transition[i]]
print " %s: %s" % (markov_model.states[i], ' '.join(x))
print "EMISSION:"
for i in range(len(markov_model.p_emission)):
x = ["%.2f" % x for x in markov_model.p_emission[i]]
print " %s: %s" % (markov_model.states[i], ' '.join(x))
print "TESTING train_visible"
states = ["0", "1", "2", "3"]
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")
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]
p_transition = [[0.7, 0.3],
[0.5, 0.5]]
p_emission = [[0.6, 0.1, 0.3],
[0.1, 0.7, 0.2]]
N, M = len(states), len(alphabet)
x = MarkovModel._baum_welch(N, M, outputs,
p_initial=p_initial,
p_transition=p_transition,
p_emission=p_emission
)
p_initial, p_transition, p_emission = x
mm = MarkovModel.MarkovModel(states, alphabet,
p_initial, p_transition, p_emission)
print_mm(mm)
# Test Baum-Welch. This is hard because it is a non-deterministic
# algorithm. Each run will result in different states having to
# different emissions. In order to help this, we need to specify some
# initial probabilities to bias the final results. This is not
# implemented yet in the MarkovModel module.
## states = [
## "state0",
## "state1",
## "state2",
## "state3",
## ]
## alphabet = ["a", "c", "g", "t"]
## training_data = [
## "aacccgggttttttt",
## "accgttttttt",
## "acgggtttttt",
## "accgtttttttt",
## "aaccgtttttttt",
## "aacggttttt",
## "acccggttttt",
## "acccggggttt",
## "aacccggggtttt",
## "aaccgggtttttt"
## ]
## print "TRAINING HMM"
## ep = ones((len(states), len(alphabet)))
## hmm = MarkovModel.train_bw(states, alphabet, training_data,
## pseudo_emission=ep)
## print "CLASSIFYING"
## states = MarkovModel.find_states(hmm, "aacgtt")
## print states
## print "STATES: %s" % ' '.join(hmm.states)
## print "ALPHABET: %s" % ' '.join(hmm.alphabet)
## print "INITIAL:"
## for i in range(len(hmm.p_initial)):
## print " %s: %.2f" % (hmm.states[i], hmm.p_initial[i])
## print "TRANSITION:"
## for i in range(len(hmm.p_transition)):
## x = ["%.2f" % x for x in hmm.p_transition[i]]
## print " %s: %s" % (hmm.states[i], ' '.join(x))
## print "EMISSION:"
## for i in range(len(hmm.p_emission)):
## x = ["%.2f" % x for x in hmm.p_emission[i]]
## print " %s: %s" % (hmm.states[i], ' '.join(x))
# Do some tests from the topcoder competition.
class DNAStrand:
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)
ds = DNAStrand()
# NNNN
print ds.mostLikely([30, 20, 30, 20, 10],
[10, 40, 10, 40, 20],
"TGCC"
)
# NNNRRRNNRRNRRN
print ds.mostLikely([4, 14, 62, 20, 44],
[39, 15, 4, 42, 25],
"CCTGAGTTAGTCGT"
)
# NRRRRRRRRRRRNNNNRRRRRRRRR
print ds.mostLikely([45, 36, 6, 13, 25],
[24, 18, 12, 46, 25],
"CCGTACTTACCCAGGACCGCAGTCC"
)
# NRRRRRRRRRR
print ds.mostLikely([75,3,1,21,45],
[34,11,39,16,15],
"TTAGCAGTGCG"
)
# N
print ds.mostLikely([26,37,8,29,16],
[31,13,33,23,25],
"T"
)
|
