def _check_and_straighten_at_triple_bond(at, bond, conf):
if at.GetDegree() != 2:
raise ValueError("only works with degree 2")
nbrs = [x.GetIdx() for x in at.GetNeighbors()]
angle = rdMolTransforms.GetAngleRad(conf, nbrs[0], at.GetIdx(), nbrs[1])
# are we off by more than a degree?
if (abs(abs(angle) - math.pi) > 0.017):
rdMolTransforms.SetAngleRad(conf, nbrs[0], at.GetIdx(), nbrs[1],
math.pi)
def _cleanup_allenes(m):
conf = m.GetConformer()
if conf.Is3D():
raise ValueError("can only operate on 2D conformers")
p = Chem.MolFromSmarts('*=[C;R0]=*')
for match in m.GetSubstructMatches(p):
angle = rdMolTransforms.GetAngleRad(conf, match[0], match[1], match[2])
# are we off by more than a degree?
if (abs(abs(angle) - math.pi) > 0.017):
rdMolTransforms.SetAngleRad(conf, match[0], match[1], match[2],
math.pi)
def testGetSetAngle(self):
file = os.path.join(RDConfig.RDBaseDir, 'Code', 'GraphMol',
'MolTransforms', 'test_data',
'3-cyclohexylpyridine.mol')
m = Chem.MolFromMolFile(file, True, False)
conf = m.GetConformer()
angle = rdmt.GetAngleDeg(conf, 0, 19, 21)
self.failUnlessAlmostEqual(angle, 109.7, 1)
rdmt.SetAngleDeg(conf, 0, 19, 21, 125.0)
angle = rdmt.GetAngleDeg(conf, 0, 19, 21)
self.failUnlessAlmostEqual(angle, 125.0, 1)
rdmt.SetAngleRad(conf, 21, 19, 0, math.pi / 2.)
angle = rdmt.GetAngleRad(conf, 0, 19, 21)
self.failUnlessAlmostEqual(angle, math.pi / 2., 1)
angle = rdmt.GetAngleDeg(conf, 0, 19, 21)
self.failUnlessAlmostEqual(angle, 90.0, 1)
def GetRingBondAng(mol, ringpath):
"""
Get bond angles of the ring
Input:
mol: rdmol
ringpath: list
Return:
bondang: list (output in radian)
"""
N = len(ringpath)
atoms = [[ringpath[i], ringpath[(i + 1) % N], ringpath[(i + 2) % N]]
for i in range(N)]
molconf = mol.GetConformer()
bondang = [
rdMolTransforms.GetAngleRad(molconf, x[0], x[1], x[2]) for x in atoms
]
return bondang
def get_3body(self):
"""
3-body parts: angles spanned by 3 adjacent atoms,
must be a valid angle in forcefield
"""
for aj in self.m.GetAtoms():
j = aj.GetIdx()
zj = self.zs[j]
neibs = aj.GetNeighbors()
nneib = len(neibs)
if zj > 1 and nneib > 1:
for i0 in range(nneib):
for k0 in range(i0 + 1, nneib):
i, k = neibs[i0].GetIdx(), neibs[k0].GetIdx()
ias = [i, j, k]
if self.zs[i] > self.zs[k]: ias = [k, j, i]
zsi = [self.zs[ia] for ia in ias]
type3 = '-'.join(['%d' % zi for zi in zsi])
if self.imbt:
if type3 not in self.mbs3: self.mbs3.append(type3)
continue
#assert type3 in self.dic3.keys()
_theta = rdMolTransforms.GetAngleRad(
self.m.GetConformer(), ias[0], ias[1], ias[2])
theta = _theta * (-1.) if _theta < 0. else _theta
#assert theta <= np.pi
if theta > np.pi:
raise '#ERROR: `thea > np.pi?'
self.dic3[type3] += [theta]
if not self.imbt:
distr3 = []
for mb3 in self.mbs3:
self.gaussian(self.xs3, self.dic3[mb3], self.sigmas[1])
distr3.append(self.ys)
self.distr3 = distr3
def make_graph(molecule_name, gb_structure, gb_scalar_coupling, mc, ob_mc,
scc_mean_and_std):
#https://stackoverflow.com/questions/14734533/how-to-access-pandas-groupby-dataframe-by-key
#----
df = gb_scalar_coupling.get_group(molecule_name)
# ['id', 'molecule_name', 'atom_index_0', 'atom_index_1', 'type',
# 'scalar_coupling_constant', 'fc', 'sd', 'pso', 'dso'],
# nomalize
scc_arr = df[['fc', 'sd', 'pso', 'dso']].values
norm_scc = np.zeros((len(df), 4))
for i, (mean_v, std_v) in enumerate(scc_mean_and_std):
norm_scc[:, i] = (scc_arr[:, i] - mean_v) / std_v
# make coupling
coupling = (
df.id.values,
scc_arr,
df[['atom_index_0', 'atom_index_1']].values,
#type = np.array([ one_hot_encoding(t,COUPLING_TYPE) for t in df.type.values ], np.uint8)
np.array([COUPLING_TYPE.index(t) for t in df.type.values], np.int32),
df.scalar_coupling_constant.values,
norm_scc)
#----
df = gb_structure.get_group(molecule_name)
df = df.sort_values(['atom_index'], ascending=True)
# ['molecule_name', 'atom_index', 'atom', 'x', 'y', 'z']
a = df.atom.values.tolist()
xyz = df[['x', 'y', 'z']].values
mol = mol_from_axyz(a, xyz)
#---
assert ( #check
a == [
mol.GetAtomWithIdx(i).GetSymbol() for i in range(mol.GetNumAtoms())
])
#---
factory = ChemicalFeatures.BuildFeatureFactory(
os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef'))
feature = factory.GetFeaturesForMol(mol)
## ** node **
#[ a.GetSymbol() for a in mol.GetAtoms() ]
num_atom = mol.GetNumAtoms()
symbol = np.zeros((num_atom, len(SYMBOL)), np.uint8) #category
acceptor = np.zeros((num_atom, 1), np.uint8)
donor = np.zeros((num_atom, 1), np.uint8)
aromatic = np.zeros((num_atom, 1), np.uint8)
hybridization = np.zeros((num_atom, len(HYBRIDIZATION)), np.uint8)
num_h = np.zeros((num_atom, 1), np.float32) #real
atomic = np.zeros((num_atom, 1), np.float32)
# ring check
isotope = np.zeros((num_atom, 1), np.uint8)
isin_ring = np.zeros((num_atom, 1), np.uint8)
ring_types = [3, 4, 5, 6, 7, 8]
n_ring = np.zeros((num_atom, len(ring_types)), np.uint8)
for i in range(num_atom):
atom = mol.GetAtomWithIdx(i)
symbol[i] = one_hot_encoding(atom.GetSymbol(), SYMBOL)
aromatic[i] = atom.GetIsAromatic()
hybridization[i] = one_hot_encoding(atom.GetHybridization(),
HYBRIDIZATION)
num_h[i] = atom.GetTotalNumHs(includeNeighbors=True)
atomic[i] = atom.GetAtomicNum()
#new
isotope[i] = atom.GetIsotope()
isin_ring[i] = atom.IsInRing()
for j, ring_i in enumerate(ring_types):
n_ring[i, j] = atom.IsInRingSize(ring_i)
#[f.GetFamily() for f in feature]
for t in range(0, len(feature)):
if feature[t].GetFamily() == 'Donor':
for i in feature[t].GetAtomIds():
donor[i] = 1
elif feature[t].GetFamily() == 'Acceptor':
for i in feature[t].GetAtomIds():
acceptor[i] = 1
#edge_angleを出すためのindex_listの作成
n_bonds = len(mol.GetBonds())
bond_arr_0 = [(mol.GetBonds()[i].GetBeginAtomIdx(),
mol.GetBonds()[i].GetEndAtomIdx()) for i in range(n_bonds)]
bond_arr_1 = [(mol.GetBonds()[i].GetEndAtomIdx(),
mol.GetBonds()[i].GetBeginAtomIdx())
for i in range(n_bonds)]
bond_df = pd.DataFrame(bond_arr_0 + bond_arr_1, columns=["a0", "a1"])
bond_df3 = bond_df.merge(bond_df.rename(columns={
"a1": "a2",
"a0": "a1"
}),
on="a1")
bond_df3 = bond_df3[bond_df3.a0 != bond_df3.a2]
bond_df4 = bond_df3.merge(bond_df.rename(columns={
"a0": "a2",
"a1": "a3"
}),
on="a2")
bond_df4 = bond_df4[(bond_df4.a0 != bond_df4.a3)
& (bond_df4.a1 != bond_df4.a3)]
bond_idx3 = bond_df3.values
bond_idx4 = bond_df4.values
#all_bond_df = all_bond_df.sort_values(["a0","a1","a2"]).reset_index(drop=True)
## ** edge **
num_edge = num_atom * num_atom - num_atom
edge_index = np.zeros((num_edge, 2), np.uint8)
bond_type = np.zeros((num_edge, len(BOND_TYPE)), np.uint8) #category
distance = np.zeros((num_edge, 1), np.float32) #real
angle = np.zeros((num_edge, 1), np.float32) #real
dihedrals_min = np.zeros((num_edge, 1), np.float32)
dihedrals_max = np.zeros((num_edge, 1), np.float32)
dihedrals_diff = np.zeros((num_edge, 1), np.float32)
conjugate = np.zeros((num_edge, 1), np.uint8)
is_ring_edge = np.zeros((num_edge, 1), np.uint8)
bond_type_detail = np.zeros((num_edge, 1),
np.uint16) #np.zeros((num_edge,1), str)
#norm_xyz = preprocessing.normalize(xyz, norm='l2')
ij = 0
ij_dict = {}
for i in range(num_atom):
for j in range(num_atom):
if i == j: continue
edge_index[ij] = [i, j]
bond = mol.GetBondBetweenAtoms(i, j)
if bond is not None:
bond_type[ij] = one_hot_encoding(bond.GetBondType(), BOND_TYPE)
conjugate[ij] = bond.GetIsConjugated()
# add ring info
is_ring_edge[ij] = isin_ring[i, 0] and isin_ring[j, 0]
# bond type detail
#print(a[i] +"_"+ str(bond.GetBondType()) +"_" + a[j])
bond_type_detail[ij] = bond_detail_dict[a[i] + "_" + str(
bond.GetBondType()) + "_" + a[j]]
#print(a[i] + str(bond.GetBondType()) + a[j])
distance[ij] = np.linalg.norm(xyz[i] - xyz[j])
#angle[ij] = (norm_xyz[i]*norm_xyz[j]).sum()
ij_dict[(i, j)] = ij
ij += 1
conf = mol.GetConformer(0)
for i, j, k in bond_idx3:
#角Ni-Nj-Nkはi-kエッジとして格納
## なにもない0とcos = 0を区別できるように-1する
angle[ij_dict[(i, k)]] = np.cos(
rdMolTransforms.GetAngleRad(conf, int(i), int(j), int(k))) + 2
for i, j, k, l in bond_idx4:
#角Ni-Nj-Nk-Nlはi-lエッジとして格納
## なにもない0とcos = 0を区別できるように-1する
dihedral = np.cos(
rdMolTransforms.GetDihedralRad(conf, int(i), int(j), int(k),
int(l))) + 2
if angle[ij_dict[(i, k)]] < 1.1 or angle[ij_dict[(i, k)]] < 1.1:
dihedral = -1
dihedrals_min[ij_dict[(i, l)]] = min(angle[ij_dict[(i, l)]], dihedral)
dihedrals_max[ij_dict[(i,l)]]= max(angle[ij_dict[(i,l)]], dihedral)\
if angle[ij_dict[(i,l)]] != 0 else dihedral
dihedrals_diff = dihedrals_max - dihedrals_min
# add mulliken charge
mc_df = mc.get_group(molecule_name)
mc_df = mc_df.sort_values(['atom_index'], ascending=True)
mc_values = mc_df["mulliken_charge"].values.reshape([-1, 1])
ob_mc_df = ob_mc.get_group(molecule_name)
ob_mc_df = ob_mc_df.sort_values(['atom_index'], ascending=True)
ob_mc_values = ob_mc_df[[
c for c in ob_mc_df.columns
if c not in ["molecule_name", "atom_index"]
]].values
##-------------------
graph = (
molecule_name,
Chem.MolToSmiles(mol),
[a, xyz],
# node feature
[
symbol, acceptor, donor, aromatic, hybridization, num_h, isotope,
isin_ring, n_ring, atomic, mc_values, ob_mc_values
],
# edge feature
[
bond_type, distance, angle, dihedrals_min, dihedrals_max,
dihedrals_diff, conjugate, is_ring_edge
],
edge_index,
coupling,
# edge bond type detail info (string)
bond_type_detail,
)
return graph
def make_graph(molecule_name, gb_structure, gb_scalar_coupling, ):
#https://stackoverflow.com/questions/14734533/how-to-access-pandas-groupby-dataframe-by-key
#----
df = gb_scalar_coupling.get_group(molecule_name)
#----
df = gb_structure.get_group(molecule_name)
df = df.sort_values(['atom_index'], ascending=True)
a = df.atom.values.tolist()
xyz = df[['x','y','z']].values
mol = mol_from_axyz(a, xyz)
#---
assert( #check
a == [ mol.GetAtomWithIdx(i).GetSymbol() for i in range(mol.GetNumAtoms())]
)
#---
factory = ChemicalFeatures.BuildFeatureFactory(os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef'))
feature = factory.GetFeaturesForMol(mol)
## ** node **
num_atom = mol.GetNumAtoms()
symbol = np.zeros((num_atom,len(SYMBOL)),np.uint8) #category
acceptor = np.zeros((num_atom,1),np.uint8)
donor = np.zeros((num_atom,1),np.uint8)
aromatic = np.zeros((num_atom,1),np.uint8)
hybridization = np.zeros((num_atom,len(HYBRIDIZATION)),np.uint8)
num_h = np.zeros((num_atom,1),np.float32)#real
atomic = np.zeros((num_atom,1),np.float32)
# new
isotope = np.zeros((num_atom,1),np.uint8)
isin_ring = np.zeros((num_atom,1),np.uint8)
ring_types = [3,4,5,6]
n_ring = np.zeros((num_atom,len(ring_types)),np.uint8)
for i in range(num_atom):
atom = mol.GetAtomWithIdx(i)
symbol[i] = one_hot_encoding(atom.GetSymbol(),SYMBOL)
aromatic[i] = atom.GetIsAromatic()
hybridization[i] = one_hot_encoding(atom.GetHybridization(),HYBRIDIZATION)
num_h[i] = atom.GetTotalNumHs(includeNeighbors=True)
atomic[i] = atom.GetAtomicNum()
#new
isotope[i] = atom.GetIsotope()
isin_ring[i] = atom.IsInRing()
for j,ring_i in enumerate(ring_types):
n_ring[i,j] = atom.IsInRingSize(ring_i)
#[f.GetFamily() for f in feature]
for t in range(0, len(feature)):
if feature[t].GetFamily() == 'Donor':
for i in feature[t].GetAtomIds():
donor[i] = 1
elif feature[t].GetFamily() == 'Acceptor':
for i in feature[t].GetAtomIds():
acceptor[i] = 1
#edge_angleを出すためのindex_listの作成
n_bonds = len(mol.GetBonds())
bond_arr_0 = [(mol.GetBonds()[i].GetBeginAtomIdx(),
mol.GetBonds()[i].GetEndAtomIdx()) for i in range(n_bonds)]
bond_arr_1 = [(mol.GetBonds()[i].GetEndAtomIdx(),
mol.GetBonds()[i].GetBeginAtomIdx()) for i in range(n_bonds)]
bond_df = pd.DataFrame(bond_arr_0+bond_arr_1,columns=["a0","a1"])
bond_df3 = bond_df.merge(bond_df.rename(columns={"a1":"a2","a0":"a1"}), on="a1")
bond_df3 = bond_df3[bond_df3.a0 != bond_df3.a2]
bond_df4 = bond_df3.merge(bond_df.rename(columns={"a0":"a2","a1":"a3"}), on="a2")
bond_df4 = bond_df4[(bond_df4.a0 != bond_df4.a3)&(bond_df4.a1 != bond_df4.a3)]
bond_idx3 = bond_df3.values
bond_idx4 = bond_df4.values
#all_bond_df = all_bond_df.sort_values(["a0","a1","a2"]).reset_index(drop=True)
## ** edge **
num_edge = num_atom*num_atom - num_atom
edge_index = np.zeros((num_edge,2), np.uint8)
bond_type = np.zeros((num_edge,len(BOND_TYPE)), np.uint8)#category
distance = np.zeros((num_edge,1),np.float32) #real
angle = np.zeros((num_edge,1),np.float32) #real
dihedrals_min = np.zeros((num_edge,1),np.float32)
dihedrals_max = np.zeros((num_edge,1),np.float32)
dihedrals_diff = np.zeros((num_edge,1),np.float32)
conjugate = np.zeros((num_edge,1),np.uint8)
#GetIsConjugated
norm_xyz = preprocessing.normalize(xyz, norm='l2')
ij=0
ij_dict = {}
for i in range(num_atom):
for j in range(num_atom):
if i==j: continue
edge_index[ij] = [i,j]
bond = mol.GetBondBetweenAtoms(i, j)
if bond is not None:
bond_type[ij] = one_hot_encoding(bond.GetBondType(),BOND_TYPE)
conjugate[ij] = bond.GetIsConjugated()
distance[ij] = np.linalg.norm(xyz[i] - xyz[j])
#angle[ij] = (norm_xyz[i]*norm_xyz[j]).sum()
ij_dict[(i,j)] = ij
ij+=1
conf = mol.GetConformer(0)
for i,j,k in bond_idx3:
#角Ni-Nj-Nkはi-kエッジとして格納
## なにもない0とcos = 0を区別できるように-1する
angle[ij_dict[(i,k)]] = np.cos(rdMolTransforms.GetAngleRad(conf,int(i),int(j),int(k)))-1
for i,j,k,l in bond_idx4:
#角Ni-Nj-Nk-Nlはi-lエッジとして格納
## なにもない0とcos = 0を区別できるように-1する
dihedral = np.cos(rdMolTransforms.GetDihedralRad(conf,int(i),int(j),int(k),int(l)))-1
dihedrals_min[ij_dict[(i,l)]] = min(angle[ij_dict[(i,l)]], dihedral)
dihedrals_max[ij_dict[(i,l)]]= max(angle[ij_dict[(i,l)]], dihedral)\
if angle[ij_dict[(i,l)]] != 0 else dihedral
dihedrals_diff = dihedrals_max - dihedrals_min
##-------------------
atom_ret = pd.DataFrame({
"molecule_name":molecule_name,
"atom_index": df["atom_index"].values,
"is_accepter": acceptor,
"is_donor": donor,
"is_aromatic":aromatic,
"is_sp1": hybridization[:,0],
"is_sp2": hybridization[:,1],
"is_sp3": hybridization[:,2],
"is_isotope": isotope,
"isin_ring": isin_ring,
"is_ring3": n_ring[:,0],
"is_ring4": n_ring[:,1],
"is_ring5": n_ring[:,2],
"is_ring6": n_ring[:,3],
})
[bond_type, distance, angle, dihedrals_min, dihedrals_max,
dihedrals_diff, conjugate,],
edge_index,
