def charge_molecule(molecule: bytes, total_charge: int) -> str:
"""Submit a molecule for charging
Accepts input and produces output in Lemon Graph Format, for which
there is no MIME type, so this specifies text/plain (which it is).
See http://lemon.cs.elte.hu/pub/doc/1.2.3/a00002.html.
Args:
molecule: Description of the input molecule.
total_charge: Desired total charge of the molecule.
"""
try:
graph = convert_from(molecule.decode('utf-8'), IOType.LGF)
except (ValueError, AttributeError):
return ('Error decoding input, is it valid LGF, and sent as'
' "Content-Type: text/plain" ?'), 400
try:
charge_server.charge(graph, total_charge)
except AssignmentError:
return ('Charges could not be assigned due to lack of data or because'
' the total charge was too far off from our reference charges.'
), 404
lgf_output = convert_to(graph, IOType.LGF)
return lgf_output.encode()
def process(self, molid: int) -> Tuple[int, nx.Graph]:
filename = os.path.join(self.__data_location,
'%d%s' % (molid, self.__extension))
with open(filename, 'r') as f:
graph = convert_from(f.read(), self.__data_type)
if graph.number_of_nodes() == 0:
raise RuntimeError('Molecule with molid {} read from file {}'
' has no atoms! Is this file valid?'.format(
molid, filename))
return molid, graph
def test_from_rdkit(ref_graph_rdkit, ref_graph_shifted):
graph = convert_from(ref_graph_rdkit, IOType.RDKIT)
assert_same_graph(graph, ref_graph_shifted)
def test_from_itp(ref_graph_itp, ref_graph):
graph = convert_from(ref_graph_itp, IOType.ITP)
assert_same_graph(graph, ref_graph)
def test_from_gml(ref_graph_gml, ref_graph):
graph = convert_from(ref_graph_gml, IOType.GML)
assert_same_graph(graph, ref_graph)
def test_invalid_lgf():
invalid_lgf = 'This is not valid LGF'
with pytest.raises(ValueError):
graph = convert_from(invalid_lgf, IOType.LGF)
def test_from_lgf(ref_graph_lgf, ref_graph):
graph = convert_from(ref_graph_lgf, IOType.LGF)
assert_same_graph(graph, ref_graph)
def cross_validate_molecules(
charger_type: str,
iacm: bool,
data_location: str,
data_type: IOType = IOType.LGF,
shell: Union[None, int, Iterable[int]] = None,
repo: Optional[Repository] = None,
bucket: int = 0,
num_buckets: int = 1
) -> ValidationReport:
"""Cross-validates a particular method on the given molecule data.
Runs through all the molecules in the repository, and for each, \
predicts charges using the given charger type and from the rest of \
the molecules in the repository.
If iacm is False, the test molecule is stripped of its charge data \
and its IACM atom types, leaving only plain elements. It is then \
matched first against the IACM side of the repository, and if that \
yields no charges, against the plain element side of the repository.
If iacm is True, the test molecule is stripped of charges but keeps \
its IACM atom types. It is then matched against the IACM side of the \
repository, and if no matches are found, its plain elements are \
matched against the plain element side.
If bucket and num_buckets are specified, then this will only run \
the cross-validation if (molid % num_buckets) == bucket.
Args:
charger_type: Name of a Charger-derived class implementing an \
assignment method.
iacm: Whether to use IACM or plain element atoms.
data_location: Path to the directory with the molecule data.
data_type: Format of the molecule data to expect.
shell: (List of) shell size(s) to use.
repo: A Repository with traceable charges.
bucket: Cross-validate for this bucket.
num_buckets: Total number of buckets that will run.
Returns:
A dict containing AtomReports per element category, and a
MoleculeReport. Keyed by category name, and 'Molecule' for
the per-molecule statistics.
"""
if shell is None:
min_shell, max_shell = None, None
wanted_shells = None
else:
if isinstance(shell, int):
shell = [shell]
min_shell, max_shell = min(shell), max(shell)
wanted_shells = sorted(shell, reverse=True)
if repo is None:
if min_shell is not None:
repo = Repository.create_from(data_location, data_type, min_shell,
max_shell, traceable=True)
else:
repo = Repository.create_from(data_location, data_type,
traceable=True)
if wanted_shells is None:
shells = sorted(repo.charges_iacm.keys(), reverse=True)
else:
shells = []
for s in wanted_shells:
if s not in repo.charges_iacm.keys():
msg = 'Shell {} will not be used, as it is not in the repository'
warn(msg.format(s))
else:
shells.append(s)
nauty = Nauty()
extension = data_type.get_extension()
molids = [int(fn.replace(extension, ''))
for fn in os.listdir(data_location)
if fn.endswith(extension)]
report = ValidationReport()
for molid in molids:
if (molid % num_buckets) == bucket:
#print('molid: {}'.format(molid))
mol_path = os.path.join(data_location, '{}{}'.format(molid, extension))
with open(mol_path, 'r') as f:
graph = convert_from(f.read(), data_type)
report += cross_validate_molecule(repo, molid, graph, charger_type, shells, iacm, nauty)
return report