def clustering(self, smiles: list, verbose=0):
""" Clustering the smiles with Tanimoto similarity.
Args:
smiles (list): list of SMILES strings.
verbose (bool): whether showing the progress bar.
Returns:
list: SMILES clusters.
"""
self.clusters = dict()
counter = 0
it = tqdm(smiles) if verbose else smiles
for s in it:
s = Smiles(s)
if s.rdkit_mol is None:
continue
idx = self._find_similarity(s)
if idx is None:
self.clusters[counter] = [s]
counter += 1
else:
self._update_cluster(idx, s)
self.clusters = list(self.clusters.values())
return self.clusters
def create_graph(self, smi, idx, q):
try:
graph = Smiles(smi).to_graph(sparse=True)
except AttributeError:
return
fp = get_filtered_fingerprint(smi)
label = torch.tensor(list(fp), dtype=torch.long)[None, :]
q.put((graph, label, idx))
def create_graph(self, data, idx, q):
try:
graph = Smiles(data[0]).to_graph(sparse=True)
except AttributeError:
return
fp = data[1]
label = torch.tensor(fp, dtype=torch.long)[None, :]
q.put((graph, label, idx))
def smiles2graph(smiles):
try:
graph = Smiles(smiles).to_graph(sparse=True)
except AttributeError:
raise
x = torch.tensor(graph["atom_features"], dtype=torch.float)
edge_idx = graph["adjacency"].tocoo()
edge_idx = torch.tensor([edge_idx.row, edge_idx.col], dtype=torch.long)
return Data(x=x, edge_index=edge_idx)
def process(self, smiles_col, ecfp_col, label_col=None):
data_list = list()
df = pd.read_csv(self.raw_paths[0])
if label_col is None:
it = zip(df[smiles_col], df[ecfp_col])
else:
it = zip(df[smiles_col], df[ecfp_col], df[label_col])
for item in it:
smiles = item[0]
fp = item[1]
if label_col is not None:
label = item[2]
smi = Smiles(smiles)
try:
graph = smi.to_graph(sparse=True)
except AttributeError:
continue
x = torch.tensor(graph["atom_features"], dtype=torch.float)
edge_idx = graph["adjacency"].tocoo()
edge_idx = torch.tensor([edge_idx.row, edge_idx.col],
dtype=torch.long)
y = torch.tensor(list(map(int, list(fp.strip()))),
dtype=torch.long)[None, :]
if label_col is None:
data_list.append(Data(x=x, edge_index=edge_idx, y=y))
else:
data_list.append(
Data(x=x, edge_index=edge_idx, y=y, label=label))
if self.pre_filter is not None:
data_list = [data for data in data_list if self.pre_filter(data)]
if self.pre_transform is not None:
data_list = [self.pre_transform(data) for data in data_list]
data, slices = self.collate(data_list)
torch.save((data, slices), self.processed_paths[0])
def write_graphs(inpath, outpath, prefix=None):
""" Convert JAK dataset to graphs
"""
smiles = list()
labels = list()
with open(inpath, "r") as inf:
line = inf.readline()
while line:
_, sm, lb = line.strip().split(",")
if MolFromSmiles(sm) is None:
line = inf.readline()
continue
smiles.append(Smiles(sm))
labels.append(lb)
line = inf.readline()
writer = GraphWriter(smiles)
writer.write(outpath, prefix=prefix, graph_labels=labels)
def write_graphs(inpath, outpath, prefix=None):
""" Convert JAK dataset to graphs
"""
smiles = list()
fps = list()
pb = tqdm()
with open(inpath, "r") as inf:
line = inf.readline()
while line:
sm = line.strip()
if MolFromSmiles(sm) is None:
line = inf.readline()
continue
smiles.append(Smiles(sm))
fps.append(",".join(map(str, get_filtered_fingerprint(sm))))
pb.update(1)
line = inf.readline()
writer = GraphWriter(smiles)
writer.write(outpath, prefix=prefix, graph_labels=fps)
def _graph_helper(self, smi):
graph = Smiles(smi).to_graph(sparse=True)
x = torch.tensor(graph["atom_features"], dtype=torch.float)
edge_idx = graph["adjacency"].tocoo()
edge_idx = torch.tensor([edge_idx.row, edge_idx.col], dtype=torch.long)
return x, edge_idx