ml-bcell-epitope/seq-linear-prediction2-aap.py at master · johncava/ml-bcell-epitope · GitHub

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
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

import torch
import torch.autograd as autograd
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import numpy as np
from torch.autograd import Variable
from utility import *

# Vector Representations for Amino Acids
# Amino Acid: [Charge (-1/0/1), Hydropath index, Van der Waals Volume (A^3), Polar/Nonpolar(1/-1)]
table_hot = {'A': [0.0,1.8,67.0,-1.0],
             'R': [1.0,-4.5,148.0,1.0],
             'N': [0.0,-3.5,96.0,1.0],
             'D': [-1.0,-3.5,91.0,1.0],
             'C': [0.0,2.5,86.0,-1.0],
             'E': [-1.0,-3.5,109.0,1.0],
             'Q': [0.0,-3.5,114.0,1.0],
             'G': [0.0,-0.4,48.0,-1.0],
             'H': [1.0,-3.2,118.0,1.0],
             'I': [0.0,4.5,124.0,-1.0],
             'L': [0.0,3.8,124.0,-1.0],
             'K': [1.0,-3.9,135.0,1.0],
             'M': [0.0,1.9,124.0,-1.0],
             'F': [0.0,2.8,135.0,-1.0],
             'P': [0.0,-1.6,90.0,-1.0],
             'S': [0.0,-0.8,73.0,1.0],
             'T': [0.0,-0.7,93.0,1.0],
             'W': [0.0,-0.9,163.0,-1.0],
             'Y': [0.0,-1.3,141.0,1.0],
             'V': [0.0,4.2,105.0,-1.0]
             }

# load data
data = load_test('SEQ194.txt')
split = int(len(data)*0.80)
train, test = data[:split], data[split:]
print "Dataset created", len(train)

class Model(nn.Module):
	def __init__(self):
		super(Model, self).__init__()
		self.lstm = nn.LSTM(4,2)
		self.lstm2 = nn.LSTM(2,2)
		self.linear = nn.Linear(4,4)
		self.linear2 = nn.Linear(4,4)
		self.linear3 = nn.Linear(2,2)
		self.hidden = self.init_hidden()
		self.hidden2 = self.init_hidden()

	def init_hidden(self):
		return (autograd.Variable(torch.zeros(1, 1, 2)),
			autograd.Variable(torch.zeros((1, 1, 2))))

	def forward(self,i):
		i = self.linear(i)
		i = self.linear(i)
		out, self.hidden = self.lstm(i.view(1, 1, -1), self.hidden)
		out2, self.hidden2 = self.lstm2(out.view(1,1,-1), self.hidden2)
		out2 = self.linear3(out2)
		out2 = F.softmax(out2)
		return out2

def encode_input(x):
	return table_hot[x]

def encode_output(y):
	if y == '0':
		return [1.0,0.0]
	elif y == '1':
		return [0.0,1.0]

model = Model()
'''
loss_function = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=1e-3)

loss = 0

loss_array = []

for epoch in xrange(3):
	#l = 0
	# Note: reset loss such that doesn't accumulate after each epoch
	for sequence in xrange(len(train)):
		#print train[sequence][0]
		inputs = [Variable(torch.Tensor(encode_input(x))) for x in train[sequence][1]]
		outputs = [Variable(torch.Tensor(encode_output(y))).view(1,2).long() for y in train[sequence][2]]
		loss = 0
		optimizer.zero_grad()
		model.hidden = model.init_hidden()
		model.hidden2 = model.init_hidden()
		for i, label in zip(inputs,outputs):
			# Step through the sequence one element at a time.
			# after each step, hidden contains the hidden state.
			out = model(i)
			#loss += loss_function(out.view(1,2),label)
			loss += loss_function(out.view(1,2), torch.max(label, 1)[1])
			#l = loss
		loss_array.append(loss[0].data.numpy().tolist())
		#print loss_array[-1]
		#print 'Sequence ', (sequence + 1)
		loss.backward()#retain_graph=True)
		optimizer.step()

#np.save('lstm1_loss.npy',loss_array)

print 'Done'
torch.save(model.state_dict(), "seq-linear-2-aap.model")


plt.plot(xrange(1,len(loss_array) + 1), loss_array)
plt.xlabel('Iterations')
plt.ylabel('Cross Entropy Loss')
plt.title('Entropy Loss of Linear and LSTM (2 Layer) with AA Vectorization lr=1e-3')
plt.show()
plt.savefig('result_seq_linear_2_aap_lr=1e-3.png')

'''

# Testing

model.load_state_dict(torch.load('seq-linear-2-aap.model'))

for sequence in xrange(len(test)):
        inputs = [Variable(torch.Tensor(encode_input(x))) for x in test[sequence][1]]
        output = [Variable(torch.Tensor(encode_output(y))) for y in test[sequence][2]]
        model.hidden = model.init_hidden()
        model.hidden2 = model.init_hidden()
	accuracy = 0
        TP = 0
        FP = 0
        TN = 0
        FN = 0
        for i, label in zip(inputs, output):
                prediction = model(i).view(1,2)
                predict = torch.max(prediction,1)[1].data.numpy().tolist()[0]
                true = torch.max(label,0)[1].data.numpy().tolist()
                #if predict == 1.0:
                #       print 'Hello World!'
                #print true
                if predict == 1 and true == 1:
                        TP = TP + 1
                elif predict == 1 and true == 0:
                        FP = FP + 1
                elif predict == 0 and true == 0:
                        TN = TN + 1
                elif predict == 0 and true == 1:
                        FN = FN + 1
        sensitivity = TP/float(TP + FN)
        specificity = TN/float(FP + TN)
        print (TP,FP,TN,FN)