def __init__(self,
get_atomic_attributes,
node_attributes,
filename,
cols_to_read,
delimiter=',',
get_bond_attributes=None,
edge_attributes=None):
super(GraphDataset, self).__init__()
assert (get_bond_attributes is None) == (edge_attributes is None)
data_set = read_smiles_property_file(filename, cols_to_read, delimiter)
data = data_set[0]
target = data_set[1:]
clean_smiles, clean_idx = sanitize_smiles(data)
target = np.array(target).T
max_size = 0
for sm in clean_smiles:
mol = Chem.MolFromSmiles(sm)
if mol.GetNumAtoms() > max_size:
max_size = mol.GetNumAtoms()
self.target = target[clean_idx, :]
self.graphs = []
self.node_feature_matrix = []
self.adj_matrix = []
for sm in clean_smiles:
graph = Graph(sm, max_size, get_atomic_attributes,
get_bond_attributes)
self.node_feature_matrix.append(
graph.get_node_feature_matrix(node_attributes, max_size))
if get_bond_attributes is None:
self.adj_matrix.append(graph.adj_matrix)
else:
self.adj_matrix.append(
graph.get_edge_attr_adj_matrix(edge_attributes, max_size))
self.num_features = self.node_feature_matrix[0].shape[1]
def __init__(self,
filename,
cols_to_read,
delimiter=',',
tokens=None,
pad=True,
tokenize=True,
augment=False,
flip=True):
super(SmilesDataset, self).__init__()
self.tokenize = tokenize
data = read_smiles_property_file(filename, cols_to_read, delimiter)
smiles = data[0]
clean_smiles, clean_idx = sanitize_smiles(smiles)
if len(data) > 1:
target = np.array(data[1:], dtype='float')
target = np.array(target)
target = target.T
self.target = target[clean_idx]
else:
self.target = None
if augment:
clean_smiles, self.target = augment_smiles(clean_smiles,
self.target)
if pad:
clean_smiles, self.length = pad_sequences(clean_smiles)
tokens, self.token2idx, self.num_tokens = get_tokens(
clean_smiles, tokens)
if tokenize:
clean_smiles, self.tokens = seq2tensor(clean_smiles, tokens, flip)
self.data = clean_smiles
def __init__(self,
filename,
cols_to_read,
get_features,
delimiter=',',
return_smiles=False,
get_features_args=None):
super(FeatureDataset, self).__init__()
self.return_smiles = return_smiles
self.get_features = get_features
data = read_smiles_property_file(filename, cols_to_read, delimiter)
if len(cols_to_read) > 1:
assert len(cols_to_read) == len(data)
smiles = data[0]
target = np.array(data[1:], dtype='float')
target = target.T
num_targets = len(cols_to_read) - 1
target = target.reshape((-1, num_targets))
else:
smiles = data[0]
target = None
self.target = target
features, valid_idx, invalid_idx = get_features(
smiles, **get_features_args)
self.objects = [smiles[i] for i in valid_idx]
length = [len(sm) for sm in self.objects]
self.max_len = max(length)
self.data = features
def __init__(self, filename, cols_to_read, features, delimiter=',', tokens=None):
super(VanillaDataset, self).__init__()
data = read_smiles_property_file(filename, cols_to_read, delimiter)
smiles = data[0]
target = np.array(data[1], dtype='float')
clean_smiles, clean_idx = sanitize_smiles(smiles)
target = np.array(target)
self.target = target[clean_idx]
def from_smiles_file(cls, get_atomic_attributes, node_attributes, filename,
cols_to_read, delimiter=',', get_bond_attributes=None, edge_attributes=None):
data_set = read_smiles_property_file(filename, cols_to_read,
delimiter)
data = data_set[0]
target = np.array(data_set[1:]).squeeze()
clean_smiles, clean_idx = sanitize_smiles(data)
clean_mols = [Chem.MolFromSmiles(smiles) for smiles in clean_smiles]
clean_target = target[clean_idx]
return cls(get_atomic_attributes, node_attributes, clean_mols, clean_target,
get_bond_attributes, edge_attributes)
def __init__(self,
filename,
tokenized=False,
cols_to_read=None,
delimiter=',',
mol_tokens=None,
prot_tokens=None,
pad=True):
super(SmilesProteinDataset, self).__init__()
if not tokenized:
data = read_smiles_property_file(filename, cols_to_read, delimiter)
smiles = data[0]
proteins = np.array(data[1])
target = np.array(data[2], dtype='float')
clean_smiles, clean_idx = sanitize_smiles(smiles)
self.target = target[clean_idx]
proteins = list(proteins[clean_idx])
if pad:
clean_smiles, self.mol_lengths = pad_sequences(clean_smiles)
proteins, self.prot_lengths = pad_sequences(proteins)
self.mol_tokens, self.mol_token2idx, self.mol_num_tokens = \
get_tokens(clean_smiles, mol_tokens)
self.prot_tokens, self.prot_token2idx, self.prot_num_tokens = \
get_tokens(proteins, prot_tokens)
clean_smiles = seq2tensor(clean_smiles, self.mol_tokens)
proteins = seq2tensor(proteins, self.prot_tokens)
self.molecules = clean_smiles
self.proteins = proteins
else:
f = open(filename, 'rb')
data = pickle.load(f)
self.mol_tokens = data['smiles_tokens']
self.prot_tokens = data['proteins_tokens']
self.mol_num_tokens = len(data['smiles_tokens'])
self.prot_num_tokens = len(data['proteins_tokens'])
self.molecules = data['smiles']
self.proteins = data['proteins']
self.target = data['labels']
assert len(self.molecules) == len(self.proteins)
assert len(self.molecules) == len(self.target)
from openchem.data.siamese_data_layer import SiameseDataset
from openchem.utils.utils import identity
import torch
import torch.nn as nn
import numpy as np
from torch.optim import RMSprop, Adam
from torch.optim.lr_scheduler import ExponentialLR, StepLR
import torch.nn.functional as F
from sklearn.metrics import roc_auc_score, mean_squared_error, r2_score, f1_score
from openchem.data.utils import read_smiles_property_file
data = read_smiles_property_file(
'./benchmark_datasets/reactions/4_11_with_y2.csv',
cols_to_read=[11, 12, 14],
keep_header=False)
reactant1 = data[0]
reactant2 = data[1]
labels = np.array(data[2], dtype="float").reshape(-1, 1)
reactants = [reactant1[i] + " " + reactant2[i] for i in range(len(reactant2))]
from openchem.data.utils import get_tokens
tokens, _, _ = get_tokens(reactants)
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(reactants,
labels,
test_size=0.2,
random_state=42)
def __init__(self,
get_atomic_attributes,
node_attributes,
filename,
cols_to_read,
delimiter=',',
get_bond_attributes=None,
edge_attributes=None,
restrict_min_atoms=-1,
restrict_max_atoms=-1,
kekulize=True,
file_format="smi",
addHs=False,
has_3D=False,
allowed_atoms=None,
return_smiles=False,
**kwargs):
super(GraphDataset, self).__init__()
assert (get_bond_attributes is None) == (edge_attributes is None)
self.return_smiles = return_smiles
self.restrict_min_atoms = restrict_min_atoms
self.restrict_max_atoms = restrict_max_atoms
self.kekulize = kekulize
self.addHs = addHs
self.has_3D = has_3D
if file_format == "pickled":
data = pickle.load(open(filename, "rb"))
# this cleanup must be consistent with sanitize_smiles
mn, mx = restrict_min_atoms, restrict_max_atoms
indices = [
i for i, n in enumerate(data["num_atoms_all"])
if (n >= mn or mn < 0) and (n <= mx or mx < 0)
]
data = {
key: value[indices] if isinstance(value, np.ndarray) else
[value[i] for i in indices]
for key, value in data.items()
}
self.num_atoms_all = data["num_atoms_all"]
self.target = data["target"]
self.smiles = data["smiles"]
elif file_format == "smi":
data_set = read_smiles_property_file(filename, cols_to_read,
delimiter)
data = data_set[0]
if len(cols_to_read) == 1:
target = None
else:
target = data_set[1:]
clean_smiles, clean_idx, num_atoms, max_len = sanitize_smiles(
data,
min_atoms=restrict_min_atoms,
max_atoms=restrict_max_atoms,
return_num_atoms=True,
return_max_len=True)
self.max_len = max_len
if target is not None:
target = np.asarray(target, dtype=np.float).T
clean_smiles = [clean_smiles[i] for i in clean_idx]
num_atoms = [num_atoms[i] for i in clean_idx]
self.clean_idx = clean_idx
if target is not None:
self.target = target[clean_idx, :]
else:
self.target = None
self.smiles = clean_smiles
self.num_atoms_all = num_atoms
else:
raise NotImplementedError()
self.max_size = max(self.num_atoms_all)
self.node_attributes = node_attributes
self.edge_attributes = edge_attributes
self.get_atomic_attributes = get_atomic_attributes
self.get_bond_attributes = get_bond_attributes
from openchem.criterion.multitask_loss import MultitaskLoss
from sklearn.metrics import r2_score, mean_squared_error
from openchem.utils.utils import identity
import torch
import torch.nn as nn
import numpy as np
from torch.optim import RMSprop, Adam
from torch.optim.lr_scheduler import ExponentialLR, StepLR
import torch.nn.functional as F
from openchem.data.utils import read_smiles_property_file
data = read_smiles_property_file(
'benchmark_datasets/melt_temp/melting_data.txt',
cols_to_read=[0, 1],
delimiter='\t',
keep_header=False)
smiles = data[0][1:]
labels = np.array(data[1][1:], dtype='float').reshape(-1)
from openchem.data.utils import get_tokens
tokens, _, _ = get_tokens(smiles)
tokens = tokens + ' '
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(smiles,
labels,
test_size=0.2,
random_state=42)
from openchem.modules.mlp.openchem_mlp import OpenChemMLP
from openchem.data.smiles_data_layer import SmilesDataset
from openchem.criterion.multitask_loss import MultitaskLoss
import torch
import torch.nn as nn
import numpy as np
from torch.optim import RMSprop, Adam
from torch.optim.lr_scheduler import ExponentialLR, StepLR
import torch.nn.functional as F
from sklearn.metrics import roc_auc_score, mean_squared_error
from openchem.data.utils import read_smiles_property_file
data = read_smiles_property_file('./benchmark_datasets/tox21/tox21.csv',
cols_to_read=[13] + list(range(0,12)))
smiles = data[0]
labels = np.array(data[1:])
labels[np.where(labels=='')] = '999'
labels = labels.T
from openchem.data.utils import get_tokens
tokens, _, _ = get_tokens(smiles)
tokens = tokens + ' '
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(smiles, labels, test_size=0.2,
random_state=42)
from openchem.data.utils import save_smiles_property_file
node_attributes = {}
node_attributes['valence'] = Attribute('node', 'valence', one_hot=True, values=[1, 2, 3, 4, 5, 6])
node_attributes['charge'] = Attribute('node', 'charge', one_hot=True, values=[-1, 0, 1, 2, 3, 4])
node_attributes['hybridization'] = Attribute('node', 'hybridization',
one_hot=True, values=[0, 1, 2, 3, 4, 5, 6, 7])
node_attributes['aromatic'] = Attribute('node', 'aromatic', one_hot=True,
values=[0, 1])
node_attributes['atom_element'] = Attribute('node', 'atom_element',
one_hot=True,
values=list(range(11)))
from openchem.data.utils import read_smiles_property_file
data = read_smiles_property_file('./benchmark_datasets/logp_dataset/logP_labels.csv',
cols_to_read=[1,2], keep_header=False)
smiles = data[0]
labels = np.array(data[1]).reshape(-1, 1)
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(smiles, labels, test_size=0.2,
random_state=42)
from openchem.data.utils import save_smiles_property_file
save_smiles_property_file('./benchmark_datasets/logp_dataset/train.smi', X_train, y_train)
save_smiles_property_file('./benchmark_datasets/logp_dataset/test.smi', X_test, y_test)
train_dataset = GraphDataset(get_atomic_attributes, node_attributes,
'./benchmark_datasets/logp_dataset/train.smi',
def __init__(self,
get_atomic_attributes,
node_attributes,
filename,
cols_to_read,
delimiter=',',
get_bond_attributes=None,
edge_attributes=None,
restrict_min_atoms=-1,
restrict_max_atoms=-1,
kekulize=True,
file_format="smi",
addHs=False,
has_3D=False,
allowed_atoms=None,
return_smiles=False,
**kwargs):
super(GraphDataset, self).__init__()
assert (get_bond_attributes is None) == (edge_attributes is None)
self.return_smiles = return_smiles
self.restrict_min_atoms = restrict_min_atoms
self.restrict_max_atoms = restrict_max_atoms
self.kekulize = kekulize
self.addHs = addHs
self.has_3D = has_3D
if file_format == "pickled":
data = pickle.load(open(kwargs["pickled"], "rb"))
# this cleanup must be consistent with sanitize_smiles
mn, mx = restrict_min_atoms, restrict_max_atoms
indices = [
i for i, n in enumerate(data["num_atoms_all"])
if (n >= mn or mn < 0) and (n <= mx or mx < 0)
]
data = {
key: value[indices] if isinstance(value, np.ndarray) else
[value[i] for i in indices]
for key, value in data.items()
}
self.num_atoms_all = data["num_atoms_all"]
self.target = data["target"]
self.smiles = data["smiles"]
elif file_format == "smi":
data_set = read_smiles_property_file(filename, cols_to_read,
delimiter)
data = data_set[0]
if len(cols_to_read) == 1:
target = None
else:
target = data_set[1:]
clean_smiles, clean_idx, num_atoms, max_len = sanitize_smiles(
data,
min_atoms=restrict_min_atoms,
max_atoms=restrict_max_atoms,
return_num_atoms=True,
return_max_len=True)
self.max_len = max_len
if target is not None:
target = np.asarray(target, dtype=np.float).T
clean_smiles = [clean_smiles[i] for i in clean_idx]
num_atoms = [num_atoms[i] for i in clean_idx]
self.clean_idx = clean_idx
if target is not None:
self.target = target[clean_idx, :]
else:
self.target = None
self.smiles = clean_smiles
self.num_atoms_all = num_atoms
elif file_format == "sdf":
filenames = []
os.chdir("/home/Work/data/enamine_hll-500/")
for file in glob.glob("*.sdf"):
filenames.append(file)
self.num_atoms_all = []
smiles = []
rd_mols = []
for f in [filenames[10]]:
print(f)
supplier = Chem.SDMolSupplier(f, False, False)
n = len(supplier)
for i in range(n):
mol = supplier[i]
anum = [(a.GetAtomicNum() in allowed_atoms.keys())
for a in mol.GetAtoms()]
if sum(anum) == len(anum):
n = mol.GetNumAtoms()
x_coord = []
y_coord = []
z_coord = []
for k in range(n):
pos = mol.GetConformer().GetAtomPosition(k)
x_coord.append(pos.x)
y_coord.append(pos.y)
z_coord.append(pos.z)
if np.linalg.norm(z_coord, ord=2) > 1.0:
rd_mols.append(mol)
smiles.append(Chem.MolToSmiles(mol))
self.num_atoms_all.append(n)
self.smiles = smiles
self.rd_mols = rd_mols
self.target = np.ones(len(self.smiles))
else:
raise NotImplementedError()
self.max_size = max(self.num_atoms_all)
self.node_attributes = node_attributes
self.edge_attributes = edge_attributes
self.get_atomic_attributes = get_atomic_attributes
self.get_bond_attributes = get_bond_attributes