-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
383 lines (309 loc) · 14.3 KB
/
dataset.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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
import numpy as np
import scipy
import math
import torch
import torch.nn.functional as F
from torch.utils.data import Dataset
from pathlib import Path
import logging
import warnings
from tqdm.contrib.concurrent import process_map
import biotite.structure as struc
from biotite.structure.io.pdb import PDBFile
from torch.utils.data._utils.collate import default_collate
non_standard_to_standard = {
'2AS':'ASP', '3AH':'HIS', '5HP':'GLU', 'ACL':'ARG', 'AGM':'ARG', 'AIB':'ALA', 'ALM':'ALA', 'ALO':'THR', 'ALY':'LYS', 'ARM':'ARG',
'ASA':'ASP', 'ASB':'ASP', 'ASK':'ASP', 'ASL':'ASP', 'ASQ':'ASP', 'ASX':'ASP', 'AYA':'ALA', 'BCS':'CYS', 'BHD':'ASP', 'BMT':'THR', 'BNN':'ALA', # Added ASX => ASP
'BUC':'CYS', 'BUG':'LEU', 'C5C':'CYS', 'C6C':'CYS', 'CAS':'CYS', 'CCS':'CYS', 'CEA':'CYS', 'CGU':'GLU', 'CHG':'ALA', 'CLE':'LEU', 'CME':'CYS',
'CSD':'ALA', 'CSO':'CYS', 'CSP':'CYS', 'CSS':'CYS', 'CSW':'CYS', 'CSX':'CYS', 'CXM':'MET', 'CY1':'CYS', 'CY3':'CYS', 'CYG':'CYS',
'CYM':'CYS', 'CYQ':'CYS', 'DAH':'PHE', 'DAL':'ALA', 'DAR':'ARG', 'DAS':'ASP', 'DCY':'CYS', 'DGL':'GLU', 'DGN':'GLN', 'DHA':'ALA',
'DHI':'HIS', 'DIL':'ILE', 'DIV':'VAL', 'DLE':'LEU', 'DLY':'LYS', 'DNP':'ALA', 'DPN':'PHE', 'DPR':'PRO', 'DSN':'SER', 'DSP':'ASP',
'DTH':'THR', 'DTR':'TRP', 'DTY':'TYR', 'DVA':'VAL', 'EFC':'CYS', 'FLA':'ALA', 'FME':'MET', 'GGL':'GLU', 'GL3':'GLY', 'GLZ':'GLY',
'GMA':'GLU', 'GSC':'GLY', 'HAC':'ALA', 'HAR':'ARG', 'HIC':'HIS', 'HIP':'HIS', 'HMR':'ARG', 'HPQ':'PHE', 'HTR':'TRP', 'HYP':'PRO',
'IAS':'ASP', 'IIL':'ILE', 'IYR':'TYR', 'KCX':'LYS', 'LLP':'LYS', 'LLY':'LYS', 'LTR':'TRP', 'LYM':'LYS', 'LYZ':'LYS', 'MAA':'ALA', 'MEN':'ASN',
'MHS':'HIS', 'MIS':'SER', 'MLE':'LEU', 'MPQ':'GLY', 'MSA':'GLY', 'MSE':'MET', 'MVA':'VAL', 'NEM':'HIS', 'NEP':'HIS', 'NLE':'LEU',
'NLN':'LEU', 'NLP':'LEU', 'NMC':'GLY', 'OAS':'SER', 'OCS':'CYS', 'OMT':'MET', 'PAQ':'TYR', 'PCA':'GLU', 'PEC':'CYS', 'PHI':'PHE',
'PHL':'PHE', 'PR3':'CYS', 'PRR':'ALA', 'PTR':'TYR', 'PYL':'LYS', 'PYX':'CYS', 'SAC':'SER', 'SAR':'GLY', 'SCH':'CYS', 'SCS':'CYS', 'SCY':'CYS', 'SEC':'CYS', # Added pyrrolysine and selenocysteine
'SEL':'SER', 'SEP':'SER', 'SET':'SER', 'SHC':'CYS', 'SHR':'LYS', 'SMC':'CYS', 'SOC':'CYS', 'STY':'TYR', 'SVA':'SER', 'TIH':'ALA',
'TPL':'TRP', 'TPO':'THR', 'TPQ':'ALA', 'TRG':'LYS', 'TRO':'TRP', 'TYB':'TYR', 'TYI':'TYR', 'TYQ':'TYR', 'TYS':'TYR', 'TYY':'TYR'
}
three_to_one_letter = {'CYS': 'C', 'ASP': 'D', 'SER': 'S', 'GLN': 'Q', 'LYS': 'K',
'ILE': 'I', 'PRO': 'P', 'THR': 'T', 'PHE': 'F', 'ASN': 'N',
'GLY': 'G', 'HIS': 'H', 'LEU': 'L', 'ARG': 'R', 'TRP': 'W',
'ALA': 'A', 'VAL':'V', 'GLU': 'E', 'TYR': 'Y', 'MET': 'M', 'UNK': 'X'}
one_to_three_letter = {v:k for k,v in three_to_one_letter.items()}
letter_to_num = {'C': 4, 'D': 3, 'S': 15, 'Q': 5, 'K': 11, 'I': 9,
'P': 14, 'T': 16, 'F': 13, 'A': 0, 'G': 7, 'H': 8,
'E': 6, 'L': 10, 'R': 1, 'W': 17, 'V': 19,
'N': 2, 'Y': 18, 'M': 12, 'X': 20}
class ProteinDataset(Dataset):
def __init__(self, dataset_path, min_res_num=40, max_res_num=256, ss_constraints=True):
super().__init__()
# Ignore biotite warnings
warnings.filterwarnings("ignore", ".*elements were guessed from atom_.*")
self.min_res_num = min_res_num
self.max_res_num = max_res_num
self.ss_constraints = ss_constraints
# Load PDB files into dataset
paths = list(Path(dataset_path).iterdir())
structures = self.parse_pdb(paths)
# Remove None from self.structures
self.structures = [self.to_tensor(i) for i in structures if i is not None]
def parse_pdb(self, paths):
logging.info(f"Processing dataset of length {len(paths)}...")
data = list(process_map(self.get_features, paths, chunksize=10))
return data
def get_coarse_constraints(self, model, cb, dist_threshold=7, dmax=20, block_dropout=0.1):
# Used for splitting block secondary structures
def consecutive(data, stepsize=1):
return np.split(data, np.where(np.diff(data) != stepsize)[0] + 1)
dist_threshold_norm = (dist_threshold / dmax * 2) - 1
psea_to_index = {"a": 1, "b": 2, "c": 3}
chain_id = struc.get_chains(model)[0]
s = [psea_to_index[i] for i in struc.annotate_sse(model, chain_id)]
if len(s) != cb.shape[0]: return None, None # Shape mismatch from PSEA: TODO: Find issue
# annotate_sse is based on CA coordinates, so the shape is wrong if a CA coordinate is missing
# Fix by inserting 0 at indices where CA coordinates are missing
# ca_mask_index = (1-ca_atom_mask).nonzero()[0]
# [s.insert(i,0) for i in ca_mask_index]
s = np.array(s)
helix_mask = (s == 1)
beta_mask = (s == 2)
# Block adjacencies
helix_indices = helix_mask.nonzero()[0]
beta_indices = beta_mask.nonzero()[0]
helix_indices_split = [i for i in consecutive(helix_indices) if len(i) >= 4]
beta_indices_split = [i for i in consecutive(beta_indices) if len(i) >= 4]
helix_mask_pair = np.zeros(cb.shape)
for i in helix_indices_split:
start, end = i[0], i[-1]
helix_mask_pair[start:end, start:end] = 1
beta_mask_pair = np.zeros(cb.shape)
for i1 in beta_indices_split:
for i2 in beta_indices_split:
start1, end1 = i1[0], i1[-1]
start2, end2 = i2[0], i2[-1]
beta_mask_pair[start1:end1, start2:end2] = 1
helix_beta_indices = helix_indices_split + beta_indices_split
block_adj_mask = np.zeros(cb.shape)
for idx1, block1 in enumerate(helix_beta_indices):
for idx2, block2 in enumerate(helix_beta_indices):
if idx1 == idx2: continue
b1_start, b1_end = block1[0], block1[-1]
b2_start, b2_end = block2[0], block2[-1]
dist = cb[b1_start:b1_end, b2_start:b2_end].min()
if dist < dist_threshold_norm:
block_adj_mask[b1_start:b1_end, b2_start:b2_end] = 1
constraints = np.stack([helix_mask_pair, beta_mask_pair, block_adj_mask], axis=-1)
# Convert to string for dataloader
helix_beta_str = ','.join([f"{i[0]}:{i[-1]}" for i in helix_beta_indices])
return constraints, helix_beta_str
def get_features(self, path):
with open(path, "r") as f:
structure = PDBFile.read(f)
if structure.get_model_count() > 1: return None
struct = structure.get_structure()
if struc.get_chain_count(struct) > 1: return None
_, aa = struc.get_residues(struct)
# Replace nonstandard amino acids
for idx,a in enumerate(aa):
if a not in three_to_one_letter.keys():
aa[idx] = non_standard_to_standard.get(a, "UNK")
one_letter_aa = [three_to_one_letter[i] for i in aa]
aa_str = ''.join(one_letter_aa)
aa = [letter_to_num[i] for i in one_letter_aa]
nres = len(aa)
if nres > self.max_res_num or nres < self.min_res_num: return None
mask = np.ones(nres)
atom_mask = np.ones((nres, 3))
bb_coords = []
for res_idx, res in enumerate(struc.residue_iter(struct)):
# Find backbone + CB atoms
atom_types = res.get_annotation("atom_name")
all_coords = res.coord[0]
crd = []
for atom_idx, a in enumerate(["N", "CA", "C"]):
idx = np.where(atom_types == a)[0]
if idx.size == 0:
atom_mask[res_idx, atom_idx] = 0
# Rolling mask i-1 and i+1 since all 3 atoms are used for CB reconstruction
mask[res_idx] = 0
if res_idx != 0:
mask[res_idx-1] = 0
if res_idx != nres-1:
mask[res_idx+1] = 0
crd.append([0, 0, 0])
else:
crd.append(all_coords[idx[0]])
bb_coords.append(crd)
bb_coords = np.array(bb_coords)
coords_6d = get_coords6d(bb_coords, dmax=20.0, normalize=True)
coords_6d = np.nan_to_num(coords_6d)
padding = np.ones((nres,nres)).reshape(nres,nres,1)
if self.ss_constraints:
block_adj, helix_beta_str = self.get_coarse_constraints(struct[0], coords_6d[:, :, 0], dist_threshold=5)
if block_adj is None: return None
coords_6d = np.concatenate([coords_6d,block_adj,padding],axis=-1)
else:
coords_6d = np.concatenate([coords_6d, padding], axis=-1)
helix_beta_str = []
mask_pair = mask.reshape(1,-1) * mask.reshape(-1, 1) # N, N
coords_6d = coords_6d * mask_pair.reshape(nres,nres,1) # N, N, C
coords_6d = coords_6d.transpose(2,0,1) # C, N, N
return {
"id": path.stem,
"coords": bb_coords,
"coords_6d": coords_6d,
"aa": aa,
"aa_str": aa_str,
"mask_pair": mask_pair,
"ss_indices": helix_beta_str # Used for block dropout
}
def to_tensor(self, d): # this part is changed for helix project only.
feat_dtypes = {
"id": None,
"coords": None,#torch.float32,
"coords_6d": torch.float32,
"aa": None,#torch.long,
"aa_str": None,
"mask_pair": torch.bool,
"ss_indices": None
}
for k,v in d.items():
if feat_dtypes[k] is not None:
d[k] = torch.tensor(v).to(dtype=feat_dtypes[k])
return d
def __len__(self):
return len(self.structures)
def __getitem__(self, idx):
return self.structures[idx]
##### Functions below adapted from trRosetta https://github.com/RosettaCommons/trRosetta2/blob/main/trRosetta/coords6d.py
# calculate dihedral angles defined by 4 sets of points
def get_dihedrals(a, b, c, d):
# Ignore divide by zero errors
np.seterr(divide='ignore', invalid='ignore')
b0 = -1.0*(b - a)
b1 = c - b
b2 = d - c
b1 /= np.linalg.norm(b1, axis=-1)[:,None]
v = b0 - np.sum(b0*b1, axis=-1)[:,None]*b1
w = b2 - np.sum(b2*b1, axis=-1)[:,None]*b1
x = np.sum(v*w, axis=-1)
y = np.sum(np.cross(b1, v)*w, axis=-1)
return np.arctan2(y, x)
# calculate planar angles defined by 3 sets of points
def get_angles(a, b, c):
v = a - b
v /= np.linalg.norm(v, axis=-1)[:,None]
w = c - b
w /= np.linalg.norm(w, axis=-1)[:,None]
x = np.sum(v*w, axis=1)
return np.arccos(x)
# get 6d coordinates from x,y,z coords of N,Ca,C atoms
def get_coords6d(xyz, dmax=20.0, normalize=True):
nres = xyz.shape[0]
# three anchor atoms
N = xyz[:,0]
Ca = xyz[:,1]
C = xyz[:,2]
# recreate Cb given N,Ca,C
b = Ca - N
c = C - Ca
a = np.cross(b, c)
Cb = -0.58273431*a + 0.56802827*b - 0.54067466*c + Ca
# fast neighbors search to collect all
# Cb-Cb pairs within dmax
kdCb = scipy.spatial.cKDTree(Cb)
indices = kdCb.query_ball_tree(kdCb, dmax)
# indices of contacting residues
idx = np.array([[i,j] for i in range(len(indices)) for j in indices[i] if i != j]).T
idx0 = idx[0]
idx1 = idx[1]
# Cb-Cb distance matrix
dist6d = np.full((nres, nres), dmax).astype(float)
dist6d[idx0,idx1] = np.linalg.norm(Cb[idx1]-Cb[idx0], axis=-1)
# matrix of Ca-Cb-Cb-Ca dihedrals
omega6d = np.zeros((nres, nres))
omega6d[idx0,idx1] = get_dihedrals(Ca[idx0], Cb[idx0], Cb[idx1], Ca[idx1])
# matrix of polar coord theta
theta6d = np.zeros((nres, nres))
theta6d[idx0,idx1] = get_dihedrals(N[idx0], Ca[idx0], Cb[idx0], Cb[idx1])
# matrix of polar coord phi
phi6d = np.zeros((nres, nres))
phi6d[idx0,idx1] = get_angles(Ca[idx0], Cb[idx0], Cb[idx1])
# Normalize all features to [-1,1]
if normalize:
# [4A, 20A]
dist6d = (dist6d / dmax*2) - 1
# [-pi, pi]
omega6d = omega6d / math.pi
# [-pi, pi]
theta6d = theta6d / math.pi
# [0, pi]
phi6d = (phi6d / math.pi*2) - 1
coords_6d = np.stack([dist6d,omega6d,theta6d,phi6d],axis=-1)
return coords_6d
class PaddingCollate(object):
def __init__(self, max_len=None):
super().__init__()
self.max_len = max_len
@staticmethod
def _pad_last(x, n, value=0):
if isinstance(x, torch.Tensor):
assert x.size(0) <= n
if x.size(0) == n:
return x
# Pairwise embeddings TODO: not very elegant
if len(x.shape) >= 2 and x.shape[-1] != 3 and x.shape[-1] == x.shape[-2]:
x = F.pad(x, (0,n-x.shape[-1],0,n-x.shape[-2]), value=value)
return x
pad_size = [n - x.size(0)] + list(x.shape[1:])
pad = torch.full(pad_size, fill_value=value).to(x)
return torch.cat([x, pad], dim=0)
elif isinstance(x, str):
pad = value * (n - len(x))
return x #+ pad # only change for coding check. should change this back.
elif isinstance(x, list):
pad = [value] * (n - len(x))
return x + pad
else:
return x
@staticmethod
def _get_value(k):
if k in ["aa_str"]:
return "_"
elif k == "aa":
return 21 # masking value
elif k in ["id","ss_indices"]:
return ''
else:
return 0
def __call__(self, data_list):
max_length = self.max_len if self.max_len else max([len(data["aa"]) for data in data_list])
data_list_padded = []
for data in data_list:
data_padded = {
k: self._pad_last(v, max_length, value=self._get_value(k)) for k,v in data.items()
}
data_list_padded.append(data_padded)
return default_collate(data_list_padded)
if __name__ == "__main__":
import matplotlib.pyplot as plt
ds = ProteinDataset(
"../diffprot/data/cath",
max_res_num = 64
)
print(len(ds))
# for i in range(7):
# plt.imshow(ds[-1]["coords_6d"][:,:,i].numpy())
# plt.savefig(f"test{i}.png")
#
# dl = torch.utils.data.DataLoader(
# ds,
# batch_size=8,
# collate_fn=PaddingCollate(max_len=128),
# )
#
# batch = next(iter(dl))
# print(batch["coords_6d"].shape)
# print(batch["aa"].shape)