-
Notifications
You must be signed in to change notification settings - Fork 1
/
Data.py
230 lines (183 loc) · 7.48 KB
/
Data.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Nov 14 15:43:02 2019
@author: shuix007
"""
import dgl
import numpy as np
from ase.units import Hartree, eV
import math
import torch
from scipy.spatial import distance
from scipy.sparse import csr_matrix
attr_index = ['A', 'B', 'C', 'mu', 'alpha', 'homo', 'lumo', 'gap', 'r2', 'zpve', 'U0', 'U', 'H', 'G', 'Cv']
atom_index = {'H':0, 'C':1, 'N':2, 'O':3, 'F':4}
atom_type = ['H', 'C', 'N', 'O', 'F']
lg_node_type = []
lg_node_index = {}
for i in range(len(atom_index)):
for j in range(i, len(atom_index)):
lg_node_index[(i, j)] = len(lg_node_type)
lg_node_type.append((i, j))
def my_float(num):
''' my function to convert complicated string to float
'''
try:
return float(num)
except:
pos = num.find('*^')
base = num[:pos]
exp = num[pos+2:]
return float(base + 'e' + exp)
def setxor(a, b):
n = len(a)
res = []
link = []
i, j = 0, 0
while i < n and j < n:
if a[i] == b[j]:
link.append(a[i])
i += 1
j += 1
elif a[i] < b[j]:
res.append(a[i])
i += 1
else:
res.append(b[j])
j += 1
if i < j:
res.append(a[-1])
elif i > j:
res.append(b[-1])
else:
link.append(a[-1])
return res, link
class Molecule(object):
def __init__(self, filename, cut_r):
''' @brief Function to read in molecues
@params filename: .xyz file
cut_r: cut off distance
'''
self.cut_r = cut_r
# read in basic information
f = open(filename)
self.nattr = len(attr_index)
self.na = int(f.readline()) # number of atoms
properties = f.readline().split() # list of properties
self.id = int(properties[1])
self.properties = dict()
self.atoms = np.zeros(self.na, dtype=np.int64)
self.coordinates = np.zeros((self.na, 3), dtype=np.float32)
self.charge = np.zeros(self.na, dtype=np.float32)
for i in range(self.nattr):
self.properties[attr_index[i]] = float(properties[i + 2])
# convert Hartree to eV
self.properties['homo'] *= Hartree / eV
self.properties['lumo'] *= Hartree / eV
self.properties['gap'] *= Hartree / eV
self.properties['zpve'] *= Hartree / eV
self.properties['U0'] *= Hartree / eV
self.properties['U'] *= Hartree / eV
self.properties['H'] *= Hartree / eV
self.properties['G'] *= Hartree / eV
for i in range(self.na):
tp = f.readline().split()
self.atoms[i] = atom_index[tp[0]]
self.coordinates[i, :] = [my_float(tp[j]) for j in range(1, 4)]
self.charge[i] = my_float(tp[4])
# skip one line
tp = f.readline()
# extract the smile representation
tp = f.readline()
self.smile = tp[:tp.find('\t')]
f.close()
self.dist = distance.cdist(self.coordinates, self.coordinates, 'euclidean')
np.fill_diagonal(self.dist, np.inf)
self._build_hetero_graph()
def _build_hetero_graph(self):
################################
# build the atom to atom graph #
################################
self.dg_num_nodes = self.na
self.dg_node_feat_discrete = torch.LongTensor(self.atoms)
dist_graph_base = self.dist.copy()
self.dg_edge_feat = torch.FloatTensor(dist_graph_base[dist_graph_base < self.cut_r]).unsqueeze(1)
dist_graph_base[dist_graph_base >= self.cut_r] = 0.
atom_graph = dgl.graph(csr_matrix(dist_graph_base), 'atom', 'a2a')
################################
# build the bond to bond graph #
################################
num_atoms = self.dist.shape[0]
bond_feat_discrete = []
bond_feat_continuous = []
indices = []
for i in range(num_atoms):
for j in range(i+1, num_atoms):
a = self.dist[i, j]
if a < self.cut_r:
bond_feat_continuous.append([a])
indices.append([i, j])
tp = tuple(sorted(self.atoms[[i, j]]))
bond_feat_discrete.append(lg_node_index[tp])
num_bonds = len(indices)
self.lg_num_nodes = num_bonds
self.lg_node_feat_discrete = torch.LongTensor(bond_feat_discrete)
self.lg_node_feat_continuous = torch.FloatTensor(bond_feat_continuous)
#######################################################
# build the atom to bond graph and bond to atom graph #
#######################################################
assignment = np.zeros((num_atoms, num_bonds), dtype=np.int64)
for i, idx in enumerate(indices):
assignment[idx[0], i] = 1
assignment[idx[1], i] = 1
bipartite_graph_base = csr_matrix(assignment)
atom2bond_graph = dgl.bipartite(bipartite_graph_base, 'atom', 'a2b', 'bond')
bond2atom_graph = dgl.bipartite(bipartite_graph_base.transpose(), 'bond', 'b2a', 'atom')
################################
# build the bond to bond graph #
################################
bond_graph_base = assignment.T @ assignment
np.fill_diagonal(bond_graph_base, 0) # eliminate self connections
bond_graph = dgl.graph(csr_matrix(bond_graph_base), 'bond', 'b2b')
self.hetero_graph = dgl.hetero_from_relations([atom_graph, atom2bond_graph, bond2atom_graph, bond_graph])
##############################################
# build edge feature for the bond2bond graph #
##############################################
x, y = np.where(bond_graph_base > 0)
num_edges = len(x)
edge_feat_continuous = np.zeros_like(x, dtype=np.float32)
for i in range(num_edges):
body1 = indices[x[i]]
body2 = indices[y[i]]
bodyxor, link = setxor(body1, body2)
a = self.dist[body1[0], body1[1]]
b = self.dist[body2[0], body2[1]]
c = self.dist[bodyxor[0], bodyxor[1]]
edge_feat_continuous[i] = self._cos_formula(a, b, c) # calculate the cos value of the angle (-1, 1)
self.lg_edge_feat = torch.FloatTensor(edge_feat_continuous).unsqueeze(1)
def _cos_formula(self, a, b, c):
''' formula to calculate the angle between two edges
a and b are the edge lengths, c is the angle length.
'''
res = (a**2 + b**2 - c**2) / (2 * a * b)
# sanity check
res = -1. if res < -1. else res
res = 1. if res > 1. else res
return np.arccos(res)
def get_hetero_graph(self):
return self.hetero_graph
def get_dg_node_feat_discrete(self):
return self.dg_node_feat_discrete
def get_lg_node_feat_continuous(self):
return self.lg_node_feat_continuous
def get_lg_node_feat_discrete(self):
return self.lg_node_feat_discrete
def get_dg_edge_feat(self):
return self.dg_edge_feat
def get_lg_edge_feat(self):
return self.lg_edge_feat
def get_dg_num_nodes(self):
return self.dg_num_nodes
def get_lg_num_nodes(self):
return self.lg_num_nodes