def _generate_charges(self, molecule):
"""Generates a set of partial charges for a molecule using
the specified charge backend.
Parameters
----------
molecule: openforcefield.topology.Molecule
The molecule to assign charges to.
"""
if self.charge_backend == BuildTLeapSystem.ChargeBackend.OpenEye:
from openforcefield.utils.toolkits import OpenEyeToolkitWrapper
toolkit_wrapper = OpenEyeToolkitWrapper()
elif self.charge_backend == BuildTLeapSystem.ChargeBackend.AmberTools:
from openforcefield.utils.toolkits import (
RDKitToolkitWrapper,
AmberToolsToolkitWrapper,
ToolkitRegistry,
)
toolkit_wrapper = ToolkitRegistry(toolkit_precedence=[
RDKitToolkitWrapper, AmberToolsToolkitWrapper
])
else:
raise ValueError(f"Invalid toolkit specification.")
molecule.generate_conformers(toolkit_registry=toolkit_wrapper)
molecule.compute_partial_charges_am1bcc(
toolkit_registry=toolkit_wrapper)
def test_compute_partial_charges_net_charge(self):
"""Test OpenEyeToolkitWrapper compute_partial_charges() on a molecule with a net +1 charge"""
toolkit_registry = ToolkitRegistry(
toolkit_precedence=[AmberToolsToolkitWrapper, RDKitToolkitWrapper])
smiles = '[H]C([H])([H])[N+]([H])([H])[H]'
molecule = Molecule.from_smiles(smiles,
toolkit_registry=toolkit_registry)
molecule.generate_conformers(toolkit_registry=toolkit_registry)
with pytest.raises(NotImplementedError) as excinfo:
charge_model = 'notARealChargeModel'
molecule.compute_partial_charges(toolkit_registry=toolkit_registry
) #, charge_model=charge_model)
# TODO: Figure out why ['cm1', 'cm2'] fail
for charge_model in ['gas', 'mul', 'bcc']:
with pytest.raises(NotImplementedError) as excinfo:
molecule.compute_partial_charges(
toolkit_registry=toolkit_registry
) #, charge_model=charge_model)
charge_sum = 0 * unit.elementary_charge
for pc in molecule._partial_charges:
charge_sum += pc
assert 0.99 * unit.elementary_charge < charge_sum < 1.01 * unit.elementary_charge
# For now, I'm just testing AM1-BCC (will test more when the SMIRNOFF spec for other charges is finalized)
molecule.compute_partial_charges_am1bcc(
toolkit_registry=toolkit_registry)
charge_sum = 0 * unit.elementary_charge
for pc in molecule._partial_charges:
charge_sum += pc
assert 0.999 * unit.elementary_charge < charge_sum < 1.001 * unit.elementary_charge
def test_compute_partial_charges(self):
"""Test OpenEyeToolkitWrapper compute_partial_charges()"""
toolkit_registry = ToolkitRegistry(
toolkit_precedence=[AmberToolsToolkitWrapper, RDKitToolkitWrapper])
smiles = '[H]C([H])([H])C([H])([H])[H]'
molecule = Molecule.from_smiles(smiles,
toolkit_registry=toolkit_registry)
molecule.generate_conformers(toolkit_registry=toolkit_registry)
# TODO: Implementation of these tests is pending a decision on the API for our charge model
with pytest.raises(NotImplementedError) as excinfo:
charge_model = 'notARealChargeModel'
molecule.compute_partial_charges(toolkit_registry=toolkit_registry
) #, charge_model=charge_model)
# ['cm1', 'cm2']
for charge_model in ['gas', 'mul', 'bcc']:
with pytest.raises(NotImplementedError) as excinfo:
molecule.compute_partial_charges(
toolkit_registry=toolkit_registry
) #, charge_model=charge_model)
charge_sum = 0 * unit.elementary_charge
for pc in molecule._partial_charges:
charge_sum += pc
assert charge_sum < 0.01 * unit.elementary_charge
# For now, just test AM1-BCC while the SMIRNOFF spec for other charge models gets worked out
molecule.compute_partial_charges_am1bcc(
toolkit_registry=toolkit_registry) # , charge_model=charge_model)
charge_sum = 0 * unit.elementary_charge
for pc in molecule._partial_charges:
charge_sum += pc
assert charge_sum < 0.002 * unit.elementary_charge
def test_register_rdkit(self):
"""Test creation of toolkit registry with RDKit toolkit"""
# Test registration of RDKitToolkitWrapper
toolkit_precedence = [RDKitToolkitWrapper]
registry = ToolkitRegistry(toolkit_precedence=toolkit_precedence,
register_imported_toolkit_wrappers=False)
#registry.register_toolkit(RDKitToolkitWrapper)
assert set([type(c) for c in registry.registered_toolkits
]) == set([RDKitToolkitWrapper])
# Test ToolkitRegistry.resolve()
assert registry.resolve(
'to_smiles') == registry.registered_toolkits[0].to_smiles
# Test ToolkitRegistry.call()
smiles = '[H][C]([H])([H])[C]([H])([H])[H]'
molecule = registry.call('from_smiles', smiles)
smiles2 = registry.call('to_smiles', molecule)
assert smiles == smiles2
def make_registry(toolkit: str) -> ToolkitRegistry:
if toolkit.lower() == "openeye":
from openforcefield.utils.toolkits import OpenEyeToolkitWrapper
toolkit_registry = ToolkitRegistry(
toolkit_precedence=[OpenEyeToolkitWrapper])
elif toolkit.lower() == "rdkit":
from openforcefield.utils.toolkits import RDKitToolkitWrapper
toolkit_registry = ToolkitRegistry(
toolkit_precedence=[RDKitToolkitWrapper])
else:
from openff.cli.utils.exceptions import UnsupportedToolkitError
raise UnsupportedToolkitError(toolkit=toolkit)
# Checks later assume that this is length 1. This should be changed if
# multiple toolkits (i.e. RDKit and AmberTools) are needed at once
assert len(toolkit_registry.registered_toolkit_versions) == 1
return toolkit_registry
def test_register_ambertools(self):
"""Test creation of toolkit registry with AmberToolsToolkitWrapper and RDKitToolkitWrapper
"""
# Test registration of AmberToolsToolkitWrapper
toolkit_precedence = [AmberToolsToolkitWrapper, RDKitToolkitWrapper]
registry = ToolkitRegistry(toolkit_precedence=toolkit_precedence,
register_imported_toolkit_wrappers=False)
#registry.register_toolkit(AmberToolsToolkitWrapper)
assert set([type(c) for c in registry.registered_toolkits
]) == set([AmberToolsToolkitWrapper, RDKitToolkitWrapper])
# Test ToolkitRegistry.resolve()
registry.resolve('compute_partial_charges')
assert registry.resolve(
'compute_partial_charges'
) == registry.registered_toolkits[0].compute_partial_charges
# Test ToolkitRegistry.call()
registry.register_toolkit(RDKitToolkitWrapper)
smiles = '[H]C([H])([H])C([H])([H])[H]'
molecule = registry.call('from_smiles', smiles)
def get_conformer_energies(
molecule: str,
registry: ToolkitRegistry,
forcefield: str,
constrained: bool = False,
) -> List[Molecule]:
_enforce_dependency_version("openforcefield", "0.7.0")
file_format = molecule.split(".")[-1]
loaded_molecules = registry.call(
"from_file",
molecule,
file_format=file_format,
)
if type(loaded_molecules) is not list:
loaded_molecules = [loaded_molecules]
mols = [loaded_molecules[0]]
for mol in loaded_molecules[1:]:
if mol == mols[-1]:
for conformer in mol.conformers:
mols[-1].add_conformer(conformer)
else:
mols.append(molecule)
n_molecules = len(mols)
n_conformers = sum([mol.n_conformers for mol in mols])
print(
f"{n_molecules} unique molecule(s) loaded, with {n_conformers} total conformers"
)
ff = _get_forcefield(forcefield, constrained)
mols_with_charges = []
for mol in mols:
if mol.partial_charges is not None:
mols_with_charges.append(mol)
# This is duplicated from generate_conformers
minimized_mols = []
for mol in mols:
if mol in mols_with_charges:
mol_with_charge = [mol]
else:
mol_with_charge = []
simulation, partial_charges = _build_simulation(
molecule=mol,
forcefield=ff,
mols_with_charge=mol_with_charge,
)
mol._partial_charges = partial_charges
mol.properties["minimized against: "] = forcefield
conformer_property_keys = [
"original conformer energies (kcal/mol)",
"minimized conformer energies (kcal/mol)",
"RMSD of minimized conformers (angstrom)",
]
for prop in conformer_property_keys:
mol.properties[prop] = mol.n_conformers * [None]
for i, conformer in enumerate(mol.conformers):
simulation.context.setPositions(conformer)
pre_energy, pre_positions = _get_conformer_data(simulation)
mol.properties["original conformer energies (kcal/mol)"][
i] = pre_energy
simulation = _minimize_conformer(simulation, conformer)
min_energy, min_positions = _get_conformer_data(simulation)
mol.properties["minimized conformer energies (kcal/mol)"][
i] = min_energy
mol.conformers[i] = min_positions
rms = _get_rms_two_conformers(mol, pre_positions, min_positions)
mol.properties["RMSD of minimized conformers (angstrom)"][i] = rms
minimized_mols.append(mol)
return minimized_mols
def _topology_molecule_to_mol2(topology_molecule, file_name,
charge_backend):
"""Converts an `openforcefield.topology.TopologyMolecule` into a mol2 file,
generating a conformer and AM1BCC charges in the process.
.. todo :: This function uses non-public methods from the Open Force Field toolkit
and should be refactored when public methods become available
Parameters
----------
topology_molecule: openforcefield.topology.TopologyMolecule
The `TopologyMolecule` to write out as a mol2 file. The atom ordering in
this mol2 will be consistent with the topology ordering.
file_name: str
The filename to write to.
charge_backend: BuildTLeapSystem.ChargeBackend
The backend to use for conformer generation and partial charge
calculation.
"""
from openforcefield.topology import Molecule
from simtk import unit as simtk_unit
# Make a copy of the reference molecule so we can run conf gen / charge calc without modifying the original
reference_molecule = copy.deepcopy(
topology_molecule.reference_molecule)
if charge_backend == BuildTLeapSystem.ChargeBackend.OpenEye:
from openforcefield.utils.toolkits import OpenEyeToolkitWrapper
toolkit_wrapper = OpenEyeToolkitWrapper()
reference_molecule.generate_conformers(
toolkit_registry=toolkit_wrapper)
reference_molecule.compute_partial_charges_am1bcc(
toolkit_registry=toolkit_wrapper)
elif charge_backend == BuildTLeapSystem.ChargeBackend.AmberTools:
from openforcefield.utils.toolkits import RDKitToolkitWrapper, AmberToolsToolkitWrapper, ToolkitRegistry
toolkit_wrapper = ToolkitRegistry(toolkit_precedence=[
RDKitToolkitWrapper, AmberToolsToolkitWrapper
])
reference_molecule.generate_conformers(
toolkit_registry=toolkit_wrapper)
reference_molecule.compute_partial_charges_am1bcc(
toolkit_registry=toolkit_wrapper)
else:
raise ValueError(f'Invalid toolkit specification.')
# Get access to the parent topology, so we can look up the topology atom indices later.
topology = topology_molecule.topology
# Make and populate a new openforcefield.topology.Molecule
new_molecule = Molecule()
new_molecule.name = reference_molecule.name
# Add atoms to the new molecule in the correct order
for topology_atom in topology_molecule.atoms:
# Force the topology to cache the topology molecule start indices
topology.atom(topology_atom.topology_atom_index)
new_molecule.add_atom(topology_atom.atom.atomic_number,
topology_atom.atom.formal_charge,
topology_atom.atom.is_aromatic,
topology_atom.atom.stereochemistry,
topology_atom.atom.name)
# Add bonds to the new molecule
for topology_bond in topology_molecule.bonds:
# This is a temporary workaround to figure out what the "local" atom index of
# these atoms is. In other words it is the offset we need to apply to get the
# index if this were the only molecule in the whole Topology. We need to apply
# this offset because `new_molecule` begins its atom indexing at 0, not the
# real topology atom index (which we do know).
index_offset = topology_molecule._atom_start_topology_index
# Convert the `.atoms` generator into a list so we can access it by index
topology_atoms = list(topology_bond.atoms)
new_molecule.add_bond(
topology_atoms[0].topology_atom_index - index_offset,
topology_atoms[1].topology_atom_index - index_offset,
topology_bond.bond.bond_order,
topology_bond.bond.is_aromatic,
topology_bond.bond.stereochemistry,
)
# Transfer over existing conformers and partial charges, accounting for the
# reference/topology indexing differences
new_conformers = np.zeros((reference_molecule.n_atoms, 3))
new_charges = np.zeros(reference_molecule.n_atoms)
# Then iterate over the reference atoms, mapping their indices to the topology
# molecule's indexing system
for reference_atom_index in range(reference_molecule.n_atoms):
# We don't need to apply the offset here, since _ref_to_top_index is
# already "locally" indexed for this topology molecule
local_top_index = topology_molecule._ref_to_top_index[
reference_atom_index]
new_conformers[local_top_index, :] = reference_molecule.conformers[
0][reference_atom_index].value_in_unit(simtk_unit.angstrom)
new_charges[local_top_index] = reference_molecule.partial_charges[
reference_atom_index].value_in_unit(
simtk_unit.elementary_charge)
# Reattach the units
new_molecule.add_conformer(new_conformers * simtk_unit.angstrom)
new_molecule.partial_charges = new_charges * simtk_unit.elementary_charge
# Write the molecule
new_molecule.to_file(file_name, file_format='mol2')
def generate_conformers(
molecule: str,
registry: ToolkitRegistry,
forcefield: str,
constrained: bool = False,
prefix: Optional[str] = None,
) -> List[Molecule]:
_enforce_dependency_version("openforcefield", "0.7.1.")
ff = _get_forcefield(forcefield, constrained)
file_format = molecule.split(".")[-1]
# TODO: This may not preserve order of loading molecules in
ambiguous_stereochemistry = False
try:
raw_mols = registry.call(
"from_file",
molecule,
file_format=file_format,
)
except UndefinedStereochemistryError:
ambiguous_stereochemistry = True
raw_mols = registry.call(
"from_file",
molecule,
file_format=file_format,
allow_undefined_stereo=True,
)
# When failing to parse molecules (i.e. attempting to read MOL2 with
# RDKit, which is not supported) the toolkit can return an empty
# list instead of raising a specific exception
if raw_mols == []:
from openff.cli.utils.exceptions import MoleculeParsingError
raise MoleculeParsingError(toolkit_registry=registry,
filename=molecule)
mols = []
for i, mol in enumerate(raw_mols):
if prefix is not None:
mol.name = prefix
elif not mol.name:
mol.name = "molecule"
if len(raw_mols) > 1:
mol.name += str(i)
mols.append(mol)
mols = _collapse_conformers(mols)
# TODO: How to handle names of different stereoisomers? Just act like they're different conformers?
if ambiguous_stereochemistry:
mols_with_unpacked_stereoisomers = []
for mol in mols:
# TODO: This is a brute-force approach, it would be better to check stereo
# without needing to call enumerate_stereoisomers
stereoisomers = mol.enumerate_stereoisomers()
if stereoisomers:
for i, iso in enumerate(stereoisomers):
iso.name = mol.name + "_stereoisomer" + str(i)
mols_with_unpacked_stereoisomers.append(iso)
else:
mols_with_unpacked_stereoisomers.append(mol)
mols = mols_with_unpacked_stereoisomers
for mol in mols:
existing_conf = None
if mol.conformers is not None:
existing_conf = deepcopy(mol.conformers[0])
mol.generate_conformers(
toolkit_registry=registry,
n_conformers=100,
rms_cutoff=0.25 * unit.angstrom,
)
if existing_conf is not None:
mol.add_conformer(existing_conf)
# TODO: What happens if some molecules in a multi-molecule file have charges, others don't?
mols_with_charges = []
for mol in mols:
if mol.partial_charges is not None:
mols_with_charges.append(mol)
mols_out = []
for mol in mols:
if mol in mols_with_charges:
mol_with_charge = [mol]
else:
mol_with_charge = []
simulation, partial_charges = _build_simulation(
molecule=mol,
forcefield=ff,
mols_with_charge=mol_with_charge,
)
mol._partial_charges = partial_charges
for i, conformer in enumerate(mol.conformers):
simulation = _minimize_conformer(simulation, conformer)
energy, positions = _get_conformer_data(simulation)
mol = _reconstruct_mol_from_conformer(mol, positions)
_add_metadata_to_mol(mol, energy, registry, forcefield)
mols_out.append(mol)
mols_out = _sort_mols(mols_out)
return mols_out