-
Notifications
You must be signed in to change notification settings - Fork 0
/
embedding.py
206 lines (156 loc) · 7.1 KB
/
embedding.py
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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
import torch
import pandas as pd
from torch.utils.data import Dataset, DataLoader
from Bio.Align import substitution_matrices
import pickle as pk
# one-hot
def onehot(TCR, Epitope, Label):
alphabet = 'XARNDCQEGHILKMFPSTWYV'
char_to_int = dict((c, i) for i, c in enumerate(alphabet))
int_to_char = dict((i, c) for i, c in enumerate(alphabet))
expand = torch.Tensor([0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]).view(1,20)
expand = expand.type(torch.LongTensor)
train_TCR = torch.zeros(1,20)
train_TCR = train_TCR.type(torch.LongTensor)
train_Epi = train_TCR
# train_Label = torch.ones(5836,1)
# train_Label = train_Label.type(torch.LongTensor)
for i, t in enumerate(TCR):
t_encoded = [char_to_int[char] for char in t]
t_encoded = torch.LongTensor(t_encoded).view(1,-1)
t_all = torch.cat((t_encoded, expand),1)
t_last = t_all[0,0:20].view(1, 20)
train_TCR = torch.cat((train_TCR, t_last),0)
train_TCR = train_TCR[1:,:]
for i, e in enumerate(Epitope):
e_encoded = [char_to_int[char] for char in e]
e_encoded = torch.LongTensor(e_encoded).view(1,-1)
e_all = torch.cat((e_encoded, expand),1)
e_last = e_all[0,0:20].view(1,20)
train_Epi = torch.cat((train_Epi,e_last),0)
train_Epi = train_Epi[1:,:]
train_Label = torch.LongTensor(Label)
return train_TCR, train_Epi, train_Label
# BLOSUM62 Encoding
def BLOSUM_62(TCR, Epitope, Label, d_model):
blosum62 = substitution_matrices.load('BLOSUM62')
Label = torch.LongTensor(Label).view(-1,1)
n = Label.size()[0]
ext = list('********************')
tcr_embedding = torch.zeros(n,d_model,d_model)
epi_embedding = torch.zeros(n,d_model,d_model)
for ti, tcr in enumerate(TCR):
tcr = list(tcr)
tcr = tcr + ext
tcr = tcr[0:d_model]
# print(ti)
#print(tcr)
for i in range(d_model):
for j in range(d_model):
tcr_pair = (tcr[i],tcr[j])
if tcr_pair not in blosum62:
tcr_embedding[ti,i,j] = blosum62[(tuple(reversed(tcr_pair)))]
else:
tcr_embedding[ti,i,j] = blosum62[tcr_pair]
tcr_embedding[ti,j,i] = tcr_embedding[ti,i,j]
# print(tcr_embedding)
for ei, epi in enumerate(Epitope):
if ei == 0:
epi = list(epi)
epi = epi + ext
epi = epi[0:d_model]
s = 0
for i in range(d_model):
for j in range(d_model):
epi_pair = (epi[i],epi[j])
#print('i:',i,'j',j)
if epi_pair not in blosum62:
epi_embedding[ei,i,j] = blosum62[(tuple(reversed(epi_pair)))]
else:
epi_embedding[ei,i,j] = blosum62[epi_pair]
epi_embedding[ei,j,i] = epi_embedding[ei,i,j]
epi_embedding[ei] = epi_embedding[0]
# print(tcr_embedding)
#print(epi_embedding[ei,i,j])
return tcr_embedding, epi_embedding, Label
# Atchley
def Atchley(TCR, Epitope, Label, Length):
# 构建存储tcr与epi编码的tensor数组
aa_vec = pk.load(open('atchley.pk', 'rb'))
Label = torch.LongTensor(Label).view(-1, 1)
n = Label.size()[0]
ext = list('********************') # 用于扩增tcr与epi的长度
tcr_embedding = torch.zeros(n, Length, 6)
epi_embedding = torch.zeros(n, Length, 6)
# 计算在这里面计算！
for ti, tcr in enumerate(TCR):
tcr = tcr + ' ' * (Length - len(tcr))
for i in range(Length):
tcr_embedding[ti, i, :] = torch.from_numpy(aa_vec[tcr[i]])
for ei, epi in enumerate(Epitope):
epi = epi + ' ' * (Length - len(epi))
for i in range(Length):
epi_embedding[ti, i, :] = torch.from_numpy(aa_vec[epi[i]])
print("该数据集的总个数:" + str(ei))
return tcr_embedding[:, :, 0:5], epi_embedding[:, :, 0:5], Label
# Define new dataset
class MyDataset(Dataset):
def __init__(self, path, emb_type):
self.data = pd.read_csv(path, names = ['TCR','Epitope','Label'])
self.TCR = self.data['TCR']
self.Epitope = self.data['Epitope']
self.Label = self.data['Label']
self.emb_type = emb_type
if self.emb_type == 'onehot':
TCR, Epi, self.Label = onehot(self.TCR, self.Epitope, self.Label)
elif self.emb_type == 'BLOSUM62':
TCR, Epi, self.Label = BLOSUM_62(self.TCR, self.Epitope, self.Label, 20)
else:
TCR, Epi, self.Label = Atchley(self.TCR, self.Epitope, self.Label, 20)
self.pair = torch.cat((TCR, Epi), -1)
#self.Label = torch.LongTensor(self.data['Label'])
# self.pair = torch.cat((self.TCR, self.Epitope),1)
def __getitem__(self, index):
return self.pair[index], self.Label[index]
def __len__(self):
return torch.LongTensor(self.Label).size()[0]
def AtchleyTriple(TCR1,TCR2, Epitope, Label, Length):
# 构建存储tcr与epi编码的tensor数组
aa_vec = pk.load(open('atchley.pk', 'rb'))
Label = torch.LongTensor(Label).view(-1, 1)
n = Label.size()[0]
ext = list('********************') # 用于扩增tcr与epi的长度
tcr1_embedding = torch.zeros(n, Length, 6)
tcr2_embedding = torch.zeros(n, Length, 6)
epi_embedding = torch.zeros(n, Length, 6)
# 计算在这里面计算！
for ti, tcr in enumerate(TCR1):
tcr = tcr + ' ' * (Length - len(tcr))
for i in range(Length):
tcr1_embedding[ti, i, :] = torch.from_numpy(aa_vec[tcr[i]])
for ti, tcr in enumerate(TCR2):
tcr = tcr + ' ' * (Length - len(tcr))
for i in range(Length):
tcr1_embedding[ti, i, :] = torch.from_numpy(aa_vec[tcr[i]])
for ei, epi in enumerate(Epitope):
epi = epi + ' ' * (Length - len(epi))
for i in range(Length):
epi_embedding[ti, i, :] = torch.from_numpy(aa_vec[epi[i]])
print("该数据集的总个数:" + str(ei))
return tcr1_embedding[:, :, 0:5],tcr2_embedding[:, :, 0:5], epi_embedding[:, :, 0:5], Label
class TripleDataset(Dataset):
def __init__(self, path, emb_type):
self.data = pd.read_csv(path, names=['TCR1','TCR2', 'Epitope', 'Label'])
self.TCR1 = self.data['TCR1']
self.TCR2 = self.data['TCR2']
self.Epitope = self.data['Epitope']
self.Label = self.data['Label']
self.emb_type = emb_type
T1, T2, E, self.Label = Atchley(self.TCR1, self.TCR2, self.Epitope, self.Label, 20)
self.pair = torch.cat((T1, T2, E), -1)
# self.Label = torch.LongTensor(self.data['Label'])
# self.pair = torch.cat((self.TCR, self.Epitope),1)
def __getitem__(self, index):
return self.pair[index], self.Label[index]
def __len__(self):
return torch.LongTensor(self.Label).size()[0]