def mutagenicity_graph(filepath=None):
"""
Generate list of mutagenicity graphs.
Args:
filepath (str): Path to database. Default is None.
Returns:
ist: [labels, nodes, edge_indices, edges, atoms]
- labels (list): Mutagenity label (0,1).
- nodes (list): Atoms as Atomnumber array.
- edge_indices (list): Bond indices (i,j).
- edges (list): Bond type.
- atoms (list): Atom list as string.
"""
if filepath is None:
filepath = os.path.join(setup_user_database_directory(), "data",
"mutagen")
print("Database path:", filepath)
if not os.path.exists(os.path.join(filepath, "Mutagenicity")):
mutagenicity_download_dataset(filepath)
mutagenicity_extract_dataset(filepath)
data = mutagenicity_load(filepath)
return data
def cora_graph(filepath=None):
"""
Load and convert cora citation dataset.
Args:
filepath (str): Path to dataset. Default is None.
Returns:
list: [adj_matrix,X,labels]
- adj_matrix (sp.csr_matrix): Adjacency matrix.
- X (sp.csr_matrix): Node features.
- labels (np.array): Labels.
"""
user_default_base = setup_user_database_directory()
if filepath is None:
filepath = os.path.join(str(user_default_base), "data", "cora")
print("Database path:", filepath)
if not os.path.exists(os.path.join(filepath, "cora.npz")):
cora_download_dataset(filepath)
loader = np.load(os.path.join(filepath, "cora.npz"), allow_pickle=True)
loader = dict(loader)
data = cora_make_graph(loader)
return data
def mutag_graph(filepath=None):
"""
Get MUTAG dataset.
Args:
filepath (str): Path to dataset. Default is None.
Returns:
list: [labels, nodes, edge_indices, edges]
- labels (np.array): Labels.
- nodes (list): Node features.
- edge_indices (list): List of edge edge_indices of connections per molecule.
- edges (list): List of edge features
"""
if filepath is None:
filepath = os.path.join(setup_user_database_directory(), "data",
"mutagen")
print("Database path:", filepath)
if not os.path.exists(os.path.join(filepath, "MUTAG")):
mutag_download_dataset(filepath)
mutag_extract_dataset(filepath)
print("Making graph ...", end='', flush=True)
data = mutag_load(filepath)
print('done')
return data
def cora_graph(filepath=None):
"""
Load and convert cora citation dataset.
Args:
filepath (str): Path to dataset. Default is None.
Returns:
list: [nodes, edge_indices, labels, mapping]
- nodes(np.array): Node features of shape (2708,1434)
- edge_indices (np.array): Edge edge_indices of shape (5429,2)
- labels (np.array): Class labels of citations of shape (2708,)
- mapping (dict): Class label mapping.
"""
if filepath is None:
user_database = setup_user_database_directory()
filepath = os.path.join(str(user_database), "data", "cora")
print("Database path:", filepath)
if not os.path.exists(os.path.join(filepath, "cora.tgz")):
cora_download_dataset(filepath)
if not os.path.exists(os.path.join(filepath, "cora_lu")):
cora_extract_dataset(filepath)
data = cora_make_graph(os.path.join(filepath, "cora_lu", "cora"))
return data
def cora_graph(filepath=None):
"""
Load and convert cora citation dataset.
Args:
filepath (str): Path to dataset. Default is None.
Returns:
list: [A,X,labels]
- A (sp.csr_matrix): Adjacency matrix.
- X (sp.csr_matrix): Node features.
- labels (np.array): Labels.
"""
if filepath is None:
filepath = os.path.join(setup_user_database_directory(), "data",
"cora")
print("Database path:", filepath)
if not os.path.exists(os.path.join(filepath, "cora.tgz")):
cora_download_dataset(filepath)
if not os.path.exists(os.path.join(filepath, "cora_lu")):
cora_extract_dataset(filepath)
data = cora_make_graph(os.path.join(filepath, "cora_lu", "cora"))
return data
def qm9_graph(filepath=None,
max_distance=4,
max_neighbours=15,
gauss_distance=None,
max_mols=133885):
"""
Get list of graphs np.arrays for qm9 dataset.
Args:
filepath (str): Filepath to database.
max_distance (int): 4
max_neighbours (int): 15
gauss_distance (dict): {'GBins' : 20, 'GRange' : 4, 'GSigma' : 0.4}
max_mols (int): Maximum number of molecules to take from qm9. Default is 133885.
Returns:
list: List of graph props [labels, nodes, edges, edge_idx, gstates]
- labels: All labels of qm9
- nodes: List of atomic numbers for emebdding layer
- edges: Edgefeatures (inverse distance, gauss distance)
- edge_idx: Edge edge_indices (N,2)
- gstates: Graph states, mean moleculare weight - 7 g/mol
"""
if gauss_distance is None:
gauss_distance = {'gbins': 20, 'grange': 4, 'gsigma': 0.4}
user_database = setup_user_database_directory()
if filepath is None:
filepath = os.path.join(user_database, "data", "qm")
print("Database path:", filepath)
if not os.path.exists(os.path.join(filepath, "qm9.pickle")):
qm9_download_dataset(filepath)
qm9_extract_dataset(filepath)
qm9 = qm9_write_pickle(filepath)
qm9_remove_extracted_dataset(filepath)
else:
print("Loading qm9.pickle ...", end='', flush=True)
with open(os.path.join(filepath, "qm9.pickle"), 'rb') as f:
qm9 = pickle.load(f)
print('done')
# Make graph
print("Making graph ...", end='', flush=True)
out_graph = make_qm9_graph(qm9,
max_distance=max_distance,
max_neighbours=max_neighbours,
gauss_distance=gauss_distance,
max_mols=max_mols)
print('done')
return out_graph