def get_mol_block(self, conformer):
"""Returns the MOL file block with atoms and bonds.
Args:
conformer: dataset_pb2.Conformer
Returns:
list of strings
"""
contents = []
contents.append('\n')
contents.append('{:3d}{:3d} 0 0 0 0 0 0 0 0999 V2000\n'.format(
len(conformer.bond_topologies[0].atoms),
len(conformer.bond_topologies[0].bonds)))
for atom_type, coords in zip(
conformer.bond_topologies[0].atoms,
conformer.optimized_geometry.atom_positions):
contents.append(
'{:10.4f}{:10.4f}{:10.4f} {:s} 0 0 0 0 0 0 0 0 0 0 0 0\n'
.format(smu_utils_lib.bohr_to_angstroms(coords.x),
smu_utils_lib.bohr_to_angstroms(coords.y),
smu_utils_lib.bohr_to_angstroms(coords.z),
smu_utils_lib.ATOM_TYPE_TO_RDKIT[atom_type][0]))
for bond in conformer.bond_topologies[0].bonds:
contents.append('{:3d}{:3d}{:3d} 0\n'.format(
bond.atom_a + 1, bond.atom_b + 1, bond.bond_type))
return contents
def geom_to_angstroms(geometry):
"""Convert all the coordinates in `geometry` to Angstroms.
Args:
geometry: starting Geometry Returns New Geometry with adjusted coordinates.
Returns:
Coordinates in Angstroms.
"""
result = dataset_pb2.Geometry()
for atom in geometry.atom_positions:
new_atom = dataset_pb2.Geometry.AtomPos()
new_atom.x = smu_utils_lib.bohr_to_angstroms(atom.x)
new_atom.y = smu_utils_lib.bohr_to_angstroms(atom.y)
new_atom.z = smu_utils_lib.bohr_to_angstroms(atom.z)
result.atom_positions.append(new_atom)
return result
def extract_bond_lengths(conformer, dist_sig_digits, unbonded_max):
"""Yields quantized bond lengths.
Args:
conformer: dataset_pb2.Conformer
dist_sig_digits: number of digits after decimal point to keep
unbonded_max: maximum distance to report for unbonded pairs output atom
types are single charecters, sorted lexographically. bond_type is
dataset_pb2.BondTopology.BondType dist_sig_digits is a string (to avoid
vagaries of floating point compares)
Yields:
(atom type 1, atom type 2, bond type, quantized dist)
"""
bt = conformer.bond_topologies[0]
format_str = '{:.%df}' % dist_sig_digits
for atom_idx0, atom_idx1 in itertools.combinations(range(len(bt.atoms)),
r=2):
if (bt.atoms[atom_idx0] == dataset_pb2.BondTopology.ATOM_H
or bt.atoms[atom_idx1] == dataset_pb2.BondTopology.ATOM_H):
continue
bond_type = dataset_pb2.BondTopology.BOND_UNDEFINED
for bond in bt.bonds:
if ((bond.atom_a == atom_idx0 and bond.atom_b == atom_idx1) or
(bond.atom_a == atom_idx1 and bond.atom_b == atom_idx0)):
bond_type = bond.bond_type
break
geom = conformer.optimized_geometry
atom_pos0 = np.array([
geom.atom_positions[atom_idx0].x, geom.atom_positions[atom_idx0].y,
geom.atom_positions[atom_idx0].z
],
dtype=np.double)
atom_pos1 = np.array([
geom.atom_positions[atom_idx1].x, geom.atom_positions[atom_idx1].y,
geom.atom_positions[atom_idx1].z
],
dtype=np.double)
# The intention is the buckets are the left edge of an empricial CDF.
dist = (np.floor(
smu_utils_lib.bohr_to_angstroms(
np.linalg.norm(atom_pos0 - atom_pos1)) * 10**dist_sig_digits) /
10**dist_sig_digits)
if (bond_type == dataset_pb2.BondTopology.BOND_UNDEFINED
and dist > unbonded_max):
continue
atom_char0 = smu_utils_lib.ATOM_TYPE_TO_CHAR[bt.atoms[atom_idx0]]
atom_char1 = smu_utils_lib.ATOM_TYPE_TO_CHAR[bt.atoms[atom_idx1]]
if atom_char0 > atom_char1:
atom_char0, atom_char1 = atom_char1, atom_char0
yield atom_char0, atom_char1, bond_type, format_str.format(dist)
def distance_between_atoms(geom, a1, a2):
"""Return the distance between atoms `a1` and `a2` in `geom`.
Args:
geom:
a1:
a2:
Returns:
Distance in Angstroms.
"""
return smu_utils_lib.bohr_to_angstroms(
math.sqrt((geom.atom_positions[a1].x - geom.atom_positions[a2].x) *
(geom.atom_positions[a1].x - geom.atom_positions[a2].x) +
(geom.atom_positions[a1].y - geom.atom_positions[a2].y) *
(geom.atom_positions[a1].y - geom.atom_positions[a2].y) +
(geom.atom_positions[a1].z - geom.atom_positions[a2].z) *
(geom.atom_positions[a1].z - geom.atom_positions[a2].z)))
def extract_bond_lengths(conformer, dist_sig_digits, unbonded_max):
"""Yields quantized bond lengths.
Args:
conformer: dataset_pb2.Conformer
dist_sig_digits: number of digits after decimal point to keep
unbonded_max: maximum distance to report for unbonded pairs output atom
types are single charecters, sorted lexographically. bond_type is
dataset_pb2.BondTopology.BondType dist_sig_digits is a string (to avoid
vagaries of floating point compares)
Yields:
(atom type 1, atom type 2, bond type, quantized dist)
"""
# These are considered "major" or worse errors
if (conformer.properties.errors.status >= 8
or conformer.duplicated_by > 0):
return
bt = conformer.bond_topologies[0]
format_str = '{:.%df}' % dist_sig_digits
for atom_idx0, atom_idx1 in itertools.combinations(range(len(bt.atoms)),
r=2):
if (bt.atoms[atom_idx0] == dataset_pb2.BondTopology.ATOM_H
or bt.atoms[atom_idx1] == dataset_pb2.BondTopology.ATOM_H):
continue
# Hello huge hack. F-F creates problems for us because there is
# exactly one conformer that has an F-F bond. We can't create an
# empirical distribution out of 1 value. So we'll just drop that
# one and let the FF conformer have no detected geometries.
if (bt.atoms[atom_idx0] == dataset_pb2.BondTopology.ATOM_F
and bt.atoms[atom_idx1] == dataset_pb2.BondTopology.ATOM_F):
continue
bond_type = smu_utils_lib.get_bond_type(bt, atom_idx0, atom_idx1)
geom = conformer.optimized_geometry
atom_pos0 = np.array([
geom.atom_positions[atom_idx0].x, geom.atom_positions[atom_idx0].y,
geom.atom_positions[atom_idx0].z
],
dtype=np.double)
atom_pos1 = np.array([
geom.atom_positions[atom_idx1].x, geom.atom_positions[atom_idx1].y,
geom.atom_positions[atom_idx1].z
],
dtype=np.double)
# The intention is the buckets are the left edge of an empricial CDF.
dist = (np.floor(
smu_utils_lib.bohr_to_angstroms(
np.linalg.norm(atom_pos0 - atom_pos1)) * 10**dist_sig_digits) /
10**dist_sig_digits)
if (bond_type == dataset_pb2.BondTopology.BOND_UNDEFINED
and dist > unbonded_max):
continue
atom_char0 = smu_utils_lib.ATOM_TYPE_TO_CHAR[bt.atoms[atom_idx0]]
atom_char1 = smu_utils_lib.ATOM_TYPE_TO_CHAR[bt.atoms[atom_idx1]]
if atom_char0 > atom_char1:
atom_char0, atom_char1 = atom_char1, atom_char0
yield atom_char0, atom_char1, bond_type, format_str.format(dist)
