-
Notifications
You must be signed in to change notification settings - Fork 7
/
utils.py
201 lines (200 loc) · 7.02 KB
/
utils.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
import os
import torch
import numpy as np
import pandas as pd
import pickle as pk
import torch.nn as nn
import torch.nn.functional as F
from tqdm import tqdm,trange
from preprocess import *
from graph_construction import calcPROgraph
# prot_amino2id={
# '<pad>': 0, '</s>': 1, '<unk>': 2, 'A': 3,
# 'L': 4, 'G': 5, 'V': 6, 'S': 7,
# 'R': 8, 'E': 9, 'D': 10, 'T': 11,
# 'I': 12, 'P': 13, 'K': 14, 'F': 15,
# 'Q': 16, 'N': 17, 'Y': 18, 'M': 19,
# 'H': 20, 'W': 21, 'C': 22, 'X': 23,
# 'B': 24, 'O': 25, 'U': 26, 'Z': 27
# }
amino2id={
'<null_0>': 0, '<pad>': 1, '<eos>': 2, '<unk>': 3,
'L': 4, 'A': 5, 'G': 6, 'V': 7, 'S': 8, 'E': 9, 'R': 10,
'T': 11, 'I': 12, 'D': 13, 'P': 14, 'K': 15, 'Q': 16,
'N': 17, 'F': 18, 'Y': 19, 'M': 20, 'H': 21, 'W': 22,
'C': 23, 'X': 24, 'B': 25, 'U': 26, 'Z': 27, 'O': 28,
'.': 29, '-': 30, '<null_1>': 31, '<mask>': 32, '<cath>': 33, '<af2>': 34
}
class chain:
def __init__(self):
self.sequence=[]
self.amino=[]
self.coord=[]
self.site={}
self.date=''
self.length=0
self.adj=None
self.edge=None
self.feat=None
self.dssp=None
self.name=''
self.chain_name=''
self.protein_name=''
def add(self,amino,pos,coord):
self.sequence.append(DICT[amino])
self.amino.append(amino2id[DICT[amino]])
self.coord.append(coord)
self.site[pos]=self.length
self.length+=1
def process(self):
self.amino=torch.LongTensor(self.amino)
self.coord=torch.FloatTensor(self.coord)
self.label=torch.zeros_like(self.amino)
self.sequence=''.join(self.sequence)
def extract(self,model,device,path):
if len(self)>1024 or model is None:
return
f=lambda x:model(x.to(device).unsqueeze(0),[36])['representations'][36].squeeze(0).cpu()
with torch.no_grad():
feat=f(self.amino)
torch.save(feat,f'{path}/feat/{self.name}_esm2.ts')
def load_dssp(self,path):
dssp=torch.Tensor(np.load(f'{path}/dssp/{self.name}.npy'))
pos=np.load(f'{path}/dssp/{self.name}_pos.npy')
self.dssp=torch.Tensor([
-2.4492936e-16, -2.4492936e-16,
1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0
]).repeat(self.length,1)
self.rsa=torch.zeros(self.length)
for i in range(len(dssp)):
self.dssp[self.site[pos[i]]]=dssp[i]
if dssp[i][4]>0.15:
self.rsa[i]=1
self.rsa=self.rsa.bool()
def load_feat(self,path):
self.feat=torch.load(f'{path}/feat/{self.name}_esm2.ts')
def load_adj(self,path,self_cycle=False):
graph=torch.load(f'{path}/graph/{self.name}.graph')
self.adj=graph['adj'].to_dense()
self.edge=graph['edge'].to_dense()
if not self_cycle:
self.adj[range(len(self)),range(len(self))]=0
self.edge[range(len(self)),range(len(self))]=0
def get_adj(self,path,dseq=3,dr=10,dlong=5,k=10):
graph=calcPROgraph(self.sequence,self.coord,dseq,dr,dlong,k)
torch.save(graph,f'{path}/graph/{self.name}.graph')
def update(self,pos,amino):
if amino not in DICT.keys():
return
amino_id=amino2id[DICT[amino]]
idx=self.site.get(pos,None)
if idx is None:
for i in self.site.keys():
# print(i,pos)
if i[:len(pos)]==pos:
idx=self.site.get(i)
if amino_id==self.amino[idx]:
self.label[idx]=1
return
elif amino_id!=self.amino[idx]:
for i in self.site.keys():
if i[:len(pos)]==pos:
idx=self.site.get(i)
if amino_id==self.amino[idx]:
self.label[idx]=1
return
else:
self.label[idx]=1
def __len__(self):
return self.length
def __getitem__(self,idx):
return self.amino[idx],self.coord[idx],self.label[idx]
def collate_fn(batch):
edges = [item['edge'] for item in batch]
feats = [item['feat'] for item in batch]
labels = torch.cat([item['label'] for item in batch],0)
return feats,edges,labels
def extract_chain(root,pid,chain,force=False):
if not force and os.path.exists(f'{root}/purePDB/{pid}_{chain}.pdb'):
return True
if not os.path.exists(f'{root}/PDB/{pid}.pdb'):
retry=5
pdb=None
while retry>0:
try:
with rq.get(f'https://files.rcsb.org/download/{pid}.pdb') as f:
if f.status_code==200:
pdb=f.content
break
except:
retry-=1
continue
if pdb is None:
print(f'PDB file {pid} failed to download')
return False
with open(f'{root}/PDB/{pid}.pdb','wb') as f:
f.write(pdb)
lines=[]
with open(f'{root}/PDB/{pid}.pdb','r') as f:
for line in f:
if line[:6]=='HEADER':
lines.append(line)
if line[:6].strip()=='TER' and line[21]==chain:
lines.append(line)
break
feats=judge(line,None)
if feats is not None and feats[1]==chain:
lines.append(line)
with open(f'{root}/purePDB/{pid}_{chain}.pdb','w') as f:
for i in lines:
f.write(i)
return True
def process_chain(data,root,pid,model,device):
get_dssp(pid,root)
same={}
with open(f'{root}/purePDB/{pid}.pdb','r') as f:
for line in f:
if line[:6]=='HEADER':
date=line[50:59].strip()
data.date=date
continue
feats=judge(line,'CA')
if feats is None:
continue
amino,_,site,x,y,z=feats
if len(amino)>3:
if same.get(site) is None:
same[site]=amino[0]
if same[site]!=amino[0]:
continue
amino=amino[-3:]
data.add(amino,site,[x,y,z])
data.process()
data.get_adj(root)
data.extract(model,device,root)
return data
def initial(file,root,model=None,device='cpu',from_native_pdb=True):
df=pd.read_csv(f'{root}/{file}',header=0,index_col=0)
prefix=df.index
labels=df['Epitopes (resi_resn)']
samples=[]
with tqdm(prefix) as tbar:
for i in tbar:
tbar.set_postfix(protein=i)
if from_native_pdb:
state=extract_chain(root,i[:4],i[-1])
if not state:
continue
data=chain()
p,c=i.split('_')
data.protein_name=p
data.chain_name=c
data.name=f"{p}_{c}"
process_chain(data,root,i,model,device)
label=labels.loc[i].split(', ')
for j in label:
site,amino=j.split('_')
data.update(site,amino)
samples.append(data)
with open(f'{root}/total.pkl','wb') as f:
pk.dump(samples,f)