def update_feat_values(mol: Mol, atom_props: dict, bond_props: dict):
for atom in mol.GetAtoms():
for prop_key in atom_props.keys():
atom_props[prop_key].add(try_get_atom_feature(atom, prop_key))
for bond in mol.GetBonds():
for prop_key in bond_props.keys():
bond_props[prop_key].add(try_get_bond_feature(bond, prop_key))
def build_adjacency_matrix(molecule: Mol) -> np.ndarray:
adj_matrix = np.eye(molecule.GetNumAtoms())
for bond in molecule.GetBonds():
begin_atom = bond.GetBeginAtom().GetIdx()
end_atom = bond.GetEndAtom().GetIdx()
adj_matrix[begin_atom, end_atom] = adj_matrix[end_atom, begin_atom] = 1
return adj_matrix
def rdmol_to_data(mol: Mol):
assert mol.GetNumConformers() == 1
N = mol.GetNumAtoms()
pos = torch.tensor(mol.GetConformer(0).GetPositions(), dtype=torch.float)
atomic_number = []
aromatic = []
sp = []
sp2 = []
sp3 = []
num_hs = []
for atom in mol.GetAtoms():
atomic_number.append(atom.GetAtomicNum())
aromatic.append(1 if atom.GetIsAromatic() else 0)
hybridization = atom.GetHybridization()
sp.append(1 if hybridization == HybridizationType.SP else 0)
sp2.append(1 if hybridization == HybridizationType.SP2 else 0)
sp3.append(1 if hybridization == HybridizationType.SP3 else 0)
z = torch.tensor(atomic_number, dtype=torch.long)
row, col, edge_type = [], [], []
for bond in mol.GetBonds():
start, end = bond.GetBeginAtomIdx(), bond.GetEndAtomIdx()
row += [start, end]
col += [end, start]
edge_type += 2 * [BOND_TYPES[bond.GetBondType()]]
edge_index = torch.tensor([row, col], dtype=torch.long)
edge_type = torch.tensor(edge_type)
perm = (edge_index[0] * N + edge_index[1]).argsort()
edge_index = edge_index[:, perm]
edge_type = edge_type[perm]
row, col = edge_index
hs = (z == 1).to(torch.float)
num_hs = scatter(hs[row], col, dim_size=N).tolist()
smiles = Chem.MolToSmiles(mol)
data = Data(node_type=z,
pos=pos,
edge_index=edge_index,
edge_type=edge_type,
rdmol=copy.deepcopy(mol),
smiles=smiles)
data.nx = to_networkx(data, to_undirected=True)
return data
def build_bond_graph(molecule: Mol) -> Graph:
graph = Graph()
for bond in molecule.GetBonds():
source_index, sink_index = bond.GetBeginAtomIdx(), bond.GetEndAtomIdx()
graph.add_edge(source_index, sink_index)
return graph
def get_smallest_root_match(self, mol: Mol) -> Mol:
search_space: Set[Mol] = set(self.molecules.values())
all_idxs = set(range(0, mol.GetNumAtoms()))
included_idxs = {0}
bonds = {}
def register_bond(from_idx: int, to_idx: int):
entry = bonds.get(from_idx)
if entry is None:
entry = []
bonds[from_idx] = entry
entry.append(to_idx)
for bond in mol.GetBonds():
begin = bond.GetBeginAtomIdx()
end = bond.GetEndAtomIdx()
register_bond(begin, end)
register_bond(end, begin)
while len(included_idxs) < mol.GetNumAtoms():
frontier_permutations = reduce(
lambda perms, from_idx: perms | set(
map(
lambda to_idx: frozenset([*included_idxs, to_idx]),
filter(lambda idx: idx not in included_idxs, bonds[from_idx])
)
),
included_idxs,
set()
)
new_search_space = set()
for perm in frontier_permutations:
e_mol = Chem.EditableMol(mol)
perm_idxs = list(all_idxs - perm)
perm_idxs.sort(reverse=True)
for idx in perm_idxs:
e_mol.RemoveAtom(idx)
display('mul')
display_numbered(e_mol.GetMol())
new_search_space |= set(self.find_superstructures(e_mol.GetMol(), search_space))
included_idxs |= perm
if len(new_search_space) == 0:
return self.get_smallest_mol(list(search_space))
search_space = new_search_space
if len(search_space) < 100:
display("from mul")
for s in search_space:
display_numbered(s)
if len(frontier_permutations) > 1:
e_mol = Chem.EditableMol(mol)
perm_idxs = list(all_idxs - included_idxs)
perm_idxs.sort(reverse=True)
for idx in perm_idxs:
e_mol.RemoveAtom(idx)
display('single')
display_numbered(e_mol.GetMol())
new_search_space = set(self.find_superstructures(e_mol.GetMol(), search_space))
if len(new_search_space) == 0:
return self.get_smallest_mol(list(search_space))
search_space = new_search_space
if len(search_space) < 100:
display("from mul")
for s in search_space:
display_numbered(s)