def test_dihedral():
"""Test dihedrals"""
# Dihedrals
assert_array_almost_equal(
dihedral(np.array((1, 0, 0)), np.array((0, 0, 0)), np.array((0, 1, 0)),
np.array((1, 1, 0))), 0)
assert_array_almost_equal(
dihedral(np.array((1, 0, 0)), np.array((0, 0, 0)), np.array((0, 1, 0)),
np.array((1, 1, 1))), -45)
# Check benzene ring dihedral
mol = oddt.toolkit.readstring('smi', 'c1ccccc1')
mol.make3D()
assert abs(dihedral(*mol.coords[:4])) < 2.
def test_dihedral():
"""Test dihedrals"""
# Dihedrals
assert_array_almost_equal(
dihedral(np.array((1, 0, 0)), np.array((0, 0, 0)), np.array((0, 1, 0)),
np.array((1, 1, 0))), 0)
assert_array_almost_equal(
dihedral(np.array((1, 0, 0)), np.array((0, 0, 0)), np.array((0, 1, 0)),
np.array((1, 1, 1))), -45)
# Check benzene ring dihedral
mol = oddt.toolkit.readstring('smi', 'c1ccccc1')
mol.make3D()
assert_array_almost_equal(dihedral(mol.coords[0], mol.coords[1],
mol.coords[2], mol.coords[3]),
0,
decimal=1)
def test_dihedral():
"""Test dihedrals"""
# Dihedrals
assert_array_almost_equal(dihedral(np.array((1, 0, 0)),
np.array((0, 0, 0)),
np.array((0, 1, 0)),
np.array((1, 1, 0))), 0)
assert_array_almost_equal(dihedral(np.array((1, 0, 0)),
np.array((0, 0, 0)),
np.array((0, 1, 0)),
np.array((1, 1, 1))), -45)
# Check benzene ring dihedral
mol = oddt.toolkit.readstring('smi', 'c1ccccc1')
mol.make3D()
assert_array_almost_equal(dihedral(mol.coords[0],
mol.coords[1],
mol.coords[2],
mol.coords[3]), 0, decimal=1)
def test_spatial():
"""Test spatial computations"""
# Angles
assert_array_almost_equal(angle(np.array((1, 0, 0)),
np.array((0, 0, 0)),
np.array((0, 1, 0))), 90)
assert_array_almost_equal(angle(np.array((1, 0, 0)),
np.array((0, 0, 0)),
np.array((1, 1, 0))), 45)
mol = oddt.toolkit.readstring('smi', 'c1ccccc1')
mol.make3D()
# Check benzene ring angle
assert_array_almost_equal(angle(mol.coords[0],
mol.coords[1],
mol.coords[2]), 120, decimal=1)
# Dihedrals
assert_array_almost_equal(dihedral(np.array((1, 0, 0)),
np.array((0, 0, 0)),
np.array((0, 1, 0)),
np.array((1, 1, 0))), 0)
assert_array_almost_equal(dihedral(np.array((1, 0, 0)),
np.array((0, 0, 0)),
np.array((0, 1, 0)),
np.array((1, 1, 1))), -45)
# Check benzene ring dihedral
assert_array_almost_equal(dihedral(mol.coords[0],
mol.coords[1],
mol.coords[2],
mol.coords[3]), 0, decimal=1)
mol = oddt.toolkit.readstring('smi', 'c1ccccc1')
mol.make3D()
mol2 = mol.clone
# Test rotation
assert_almost_equal(mol2.coords, rotate(mol2.coords, np.pi, np.pi, np.pi))
# Rotate perpendicular to ring
mol2.coords = rotate(mol2.coords, 0, 0, np.pi)
# RMSD
assert_almost_equal(rmsd(mol, mol2, method=None), 2.77, decimal=1)
# Hungarian must be close to zero (RDKit is 0.3)
assert_almost_equal(rmsd(mol, mol2, method='hungarian'), 0, decimal=0)
# pick one molecule from docked poses
mols = list(oddt.toolkit.readfile('sdf', os.path.join(test_data_dir, 'data/dude/xiap/actives_docked.sdf')))
mols = list(filter(lambda x: x.title == '312335', mols))
assert_array_almost_equal([rmsd(mols[0], mol) for mol in mols[1:]],
[4.753552, 2.501487, 2.7941732, 1.1281863, 0.74440968,
1.6256877, 4.762476, 2.7167852, 2.5504358, 1.9303833,
2.6200771, 3.1741529, 3.225431, 4.7784939, 4.8035369,
7.8962774, 2.2385094, 4.8625236, 3.2036853])
assert_array_almost_equal([rmsd(mols[0], mol, method='hungarian') for mol in mols[1:]],
[2.5984519, 1.7295024, 1.1268076, 1.0285776, 0.73529714,
1.4094033, 2.5195069, 1.7449125, 1.5116163, 1.7796179,
2.6064286, 3.1576841, 3.2135022, 3.1675091, 2.7001681,
5.1263351, 2.0836117, 3.542397, 3.1873631])
def _dicts(self):
# Atoms
atom_dtype = [('id', 'int16'),
# atom info
('coords', 'float16', 3),
('charge', 'float16'),
('atomicnum', 'int8'),
('atomtype','a4'),
('hybridization', 'int8'),
('neighbors', 'float16', (4,3)), # non-H neighbors coordinates for angles (max of 6 neighbors should be enough)
# residue info
('resid', 'int16'),
('resname', 'a3'),
('isbackbone', 'bool'),
# atom properties
('isacceptor', 'bool'),
('isdonor', 'bool'),
('isdonorh', 'bool'),
('ismetal', 'bool'),
('ishydrophobe', 'bool'),
('isaromatic', 'bool'),
('isminus', 'bool'),
('isplus', 'bool'),
('ishalogen', 'bool'),
# secondary structure
('isalpha', 'bool'),
('isbeta', 'bool')
]
a = []
atom_dict = np.empty(self.OBMol.NumAtoms(), dtype=atom_dtype)
metals = [3,4,11,12,13,19,20,21,22,23,24,25,26,27,28,29,30,31,37,38,39,40,41,42,43,44,45,46,47,48,49,50,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,87,88,89,90,91,
92,93,94,95,96,97,98,99,100,101,102,103]
for i, atom in enumerate(self.atoms):
atomicnum = atom.atomicnum
# skip non-polar hydrogens for performance
# if atomicnum == 1 and atom.OBAtom.IsNonPolarHydrogen():
# continue
atomtype = typetable.Translate(atom.type) # sybyl atom type
partialcharge = atom.partialcharge
coords = atom.coords
if self.protein:
residue = pybel.Residue(atom.OBAtom.GetResidue())
else:
residue = False
# get neighbors, but only for those atoms which realy need them
neighbors = np.empty(4, dtype=[('coords', 'float16', 3),('atomicnum', 'int8')])
neighbors.fill(np.nan)
for n, nbr_atom in enumerate(atom.neighbors):
# concider raising neighbors list to 6, but must do some benchmarks
if n > 3:
break
nbr_atomicnum = nbr_atom.atomicnum
neighbors[n] = (nbr_atom.coords, nbr_atomicnum)
atom_dict[i] = (atom.idx,
coords,
partialcharge,
atomicnum,
atomtype,
atom.OBAtom.GetHyb(),
neighbors['coords'], #n_coords,
# residue info
residue.idx if residue else 0,
residue.name if residue else '',
residue.OBResidue.GetAtomProperty(atom.OBAtom, 2) if residue else False, # is backbone
# atom properties
atom.OBAtom.IsHbondAcceptor(),
atom.OBAtom.IsHbondDonor(),
atom.OBAtom.IsHbondDonorH(),
atomicnum in metals,
atomicnum == 6 and not (np.in1d(neighbors['atomicnum'], [6,1])).any(), #hydrophobe #doble negation, since nan gives False
atom.OBAtom.IsAromatic(),
atomtype in ['O3-', '02-' 'O-'], # is charged (minus)
atomtype in ['N3+', 'N2+', 'Ng+'], # is charged (plus)
atomicnum in [9,17,35,53], # is halogen?
False, # alpha
False # beta
)
if self.protein:
# Protein Residues (alpha helix and beta sheet)
res_dtype = [('id', 'int16'),
('resname', 'a3'),
('N', 'float16', 3),
('CA', 'float16', 3),
('C', 'float16', 3),
('isalpha', 'bool'),
('isbeta', 'bool')
] # N, CA, C
b = []
for residue in self.residues:
backbone = {}
for atom in residue:
if residue.OBResidue.GetAtomProperty(atom.OBAtom,1):
if atom.atomicnum == 7:
backbone['N'] = atom.coords
elif atom.atomicnum == 6:
if atom.type == 'C3':
backbone['CA'] = atom.coords
else:
backbone['C'] = atom.coords
if len(backbone.keys()) == 3:
b.append((residue.idx, residue.name, backbone['N'], backbone['CA'], backbone['C'], False, False))
res_dict = np.array(b, dtype=res_dtype)
# detect secondary structure by phi and psi angles
first = res_dict[:-1]
second = res_dict[1:]
psi = dihedral(first['N'], first['CA'], first['C'], second['N'])
phi = dihedral(first['C'], second['N'], second['CA'], second['C'])
# mark atoms belonging to alpha and beta
res_mask_alpha = np.where(((phi > -145) & (phi < -35) & (psi > -70) & (psi < 50))) # alpha
res_dict['isalpha'][res_mask_alpha] = True
for i in res_dict[res_mask_alpha]['id']:
atom_dict['isalpha'][atom_dict['resid'] == i] = True
res_mask_beta = np.where(((phi >= -180) & (phi < -40) & (psi <= 180) & (psi > 90)) | ((phi >= -180) & (phi < -70) & (psi <= -165))) # beta
res_dict['isbeta'][res_mask_beta] = True
atom_dict['isbeta'][np.in1d(atom_dict['resid'], res_dict[res_mask_beta]['id'])] = True
# Aromatic Rings
r = []
for ring in self.sssr:
if ring.IsAromatic():
path = ring._path
atom = atom_dict[atom_dict['id'] == path[0]]
coords = atom_dict[np.in1d(atom_dict['id'], path)]['coords']
centroid = coords.mean(axis=0)
# get vector perpendicular to ring
vector = np.cross(coords - np.vstack((coords[1:],coords[:1])), np.vstack((coords[1:],coords[:1])) - np.vstack((coords[2:],coords[:2]))).mean(axis=0) - centroid
r.append((centroid, vector, atom['isalpha'], atom['isbeta']))
ring_dict = np.array(r, dtype=[('centroid', 'float16', 3),('vector', 'float16', 3),('isalpha', 'bool'),('isbeta', 'bool'),])
self._atom_dict = atom_dict
self._ring_dict = ring_dict
if self.protein:
self._res_dict = res_dict
def detect_secondary_structure(res_dict):
"""Detect alpha helices and beta sheets in res_dict by phi and psi angles"""
first = res_dict[:-1]
second = res_dict[1:]
psi = dihedral(first['N'], first['CA'], first['C'], second['N'])
phi = dihedral(first['C'], second['N'], second['CA'], second['C'])
d = second['id'] - first['id']
# Alpha helices
res_mask_alpha = (
((phi > -145) & (phi < -35) & (psi > -70) & (psi < 50) & (d == 1))
) # alpha
res_mask_alpha = np.union1d(np.argwhere(res_mask_alpha),
np.argwhere(res_mask_alpha))
# Ignore groups smaller than 3
for mask_group in np.split(
res_mask_alpha,
np.argwhere(np.diff(res_mask_alpha) != 1).flatten() + 1):
if len(mask_group) >= 3:
res_dict['isalpha'][mask_group] = True
# Alpha helices have to form H-Bonds
hbond_dist_mask = np.abs(
res_dict[res_dict['isalpha']]['id'] -
res_dict[res_dict['isalpha']]['id'][:, np.newaxis]) >= 3
hbond_mask = distance(res_dict[res_dict['isalpha']]['N'],
res_dict[res_dict['isalpha']]['O']) < 3.5
p_mask = ((hbond_mask & hbond_dist_mask).any(axis=0) |
(hbond_mask & hbond_dist_mask).any(axis=1))
res_dict['isalpha'][np.argwhere(
res_dict['isalpha']).flatten()[~p_mask]] = False
# Ignore groups smaller than 3
res_mask_alpha = np.argwhere(res_dict['isalpha']).flatten()
for mask_group in np.split(
res_mask_alpha,
np.argwhere(np.diff(res_mask_alpha) != 1).flatten() + 1):
if 0 < len(mask_group) < 3:
res_dict['isalpha'][mask_group] = False
# Beta sheets
res_mask_beta = (
((phi >= -180) & (phi < -40) & (psi <= 180) & (psi > 90) & (d == 1)) |
((phi >= -180) & (phi < -70) & (psi <= -165) & (d == 1))) # beta
res_mask_beta = np.union1d(np.argwhere(res_mask_beta),
np.argwhere(res_mask_beta))
# Ignore groups smaller than 3
for mask_group in np.split(
res_mask_beta,
np.argwhere(np.diff(res_mask_beta) != 1).flatten() + 1):
if len(mask_group) >= 3:
res_dict['isbeta'][mask_group] = True
# Beta strands have to be alongside eachother
res_dist_mask = np.abs(res_dict[res_dict['isbeta']]['id'] -
res_dict[res_dict['isbeta']]['id'][:,
np.newaxis]) >= 4
hbond_mask = distance(res_dict[res_dict['isbeta']]['N'],
res_dict[res_dict['isbeta']]['O']) < 3.5
ca_mask = distance(res_dict[res_dict['isbeta']]['CA'],
res_dict[res_dict['isbeta']]['CA']) < 4.5
p_mask = ((hbond_mask & res_dist_mask).any(axis=0) |
(hbond_mask & res_dist_mask).any(axis=1) |
(ca_mask & res_dist_mask).any(axis=0))
res_dict['isbeta'][np.argwhere(
res_dict['isbeta']).flatten()[~p_mask]] = False
# Ignore groups smaller than 3
res_mask_beta = np.argwhere(res_dict['isbeta']).flatten()
for mask_group in np.split(
res_mask_beta,
np.argwhere(np.diff(res_mask_beta) != 1).flatten() + 1):
if 0 < len(mask_group) < 3:
res_dict['isbeta'][mask_group] = False
return res_dict
def detect_secondary_structure(res_dict):
"""Detect alpha helices and beta sheets in res_dict by phi and psi angles"""
first = res_dict[:-1]
second = res_dict[1:]
psi = dihedral(first['N'], first['CA'], first['C'], second['N'])
phi = dihedral(first['C'], second['N'], second['CA'], second['C'])
d = second['id'] - first['id']
# Alpha helices
res_mask_alpha = (((phi > -145) & (phi < -35) &
(psi > -70) & (psi < 50) & (d == 1))) # alpha
res_mask_alpha = np.union1d(np.argwhere(res_mask_alpha),
np.argwhere(res_mask_alpha))
# Ignore groups smaller than 3
for mask_group in np.split(res_mask_alpha, np.argwhere(np.diff(res_mask_alpha) != 1).flatten() + 1):
if len(mask_group) >= 3:
res_dict['isalpha'][mask_group] = True
# Alpha helices have to form H-Bonds
hbond_dist_mask = np.abs(res_dict[res_dict['isalpha']]['resnum'] -
res_dict[res_dict['isalpha']]['resnum'][:, np.newaxis]) >= 3
hbond_mask = distance(res_dict[res_dict['isalpha']]['N'],
res_dict[res_dict['isalpha']]['O']) < 3.5
p_mask = ((hbond_mask & hbond_dist_mask).any(axis=0) |
(hbond_mask & hbond_dist_mask).any(axis=1))
res_dict['isalpha'][np.argwhere(res_dict['isalpha']).flatten()[~p_mask]] = False
# Ignore groups smaller than 3
res_mask_alpha = np.argwhere(res_dict['isalpha']).flatten()
for mask_group in np.split(res_mask_alpha, np.argwhere(np.diff(res_mask_alpha) != 1).flatten() + 1):
if 0 < len(mask_group) < 3:
res_dict['isalpha'][mask_group] = False
# Beta sheets
res_mask_beta = (((phi >= -180) & (phi < -40) &
(psi <= 180) & (psi > 90) & (d == 1)) |
((phi >= -180) & (phi < -70) &
(psi <= -165) & (d == 1))) # beta
res_mask_beta = np.union1d(np.argwhere(res_mask_beta),
np.argwhere(res_mask_beta))
# Ignore groups smaller than 3
for mask_group in np.split(res_mask_beta, np.argwhere(np.diff(res_mask_beta) != 1).flatten() + 1):
if len(mask_group) >= 3:
res_dict['isbeta'][mask_group] = True
# Beta strands have to be alongside eachother
res_dist_mask = np.abs(res_dict[res_dict['isbeta']]['resnum'] -
res_dict[res_dict['isbeta']]['resnum'][:, np.newaxis]) >= 4
hbond_mask = distance(res_dict[res_dict['isbeta']]['N'],
res_dict[res_dict['isbeta']]['O']) < 3.5
ca_mask = distance(res_dict[res_dict['isbeta']]['CA'],
res_dict[res_dict['isbeta']]['CA']) < 4.5
p_mask = ((hbond_mask & res_dist_mask).any(axis=0) |
(hbond_mask & res_dist_mask).any(axis=1) |
(ca_mask & res_dist_mask).any(axis=0))
res_dict['isbeta'][np.argwhere(res_dict['isbeta']).flatten()[~p_mask]] = False
# Ignore groups smaller than 3
res_mask_beta = np.argwhere(res_dict['isbeta']).flatten()
for mask_group in np.split(res_mask_beta, np.argwhere(np.diff(res_mask_beta) != 1).flatten() + 1):
if 0 < len(mask_group) < 3:
res_dict['isbeta'][mask_group] = False
return res_dict
def _dicts(self):
# Atoms
atom_dtype = [('id', 'int16'),
# atom info
('coords', 'float32', 3),
('radius', 'float32'),
('charge', 'float32'),
('atomicnum', 'int8'),
('atomtype','a4'),
('hybridization', 'int8'),
('neighbors', 'float32', (4,3)), # non-H neighbors coordinates for angles (max of 6 neighbors should be enough)
# residue info
('resid', 'int16'),
('resname', 'a3'),
('isbackbone', 'bool'),
# atom properties
('isacceptor', 'bool'),
('isdonor', 'bool'),
('isdonorh', 'bool'),
('ismetal', 'bool'),
('ishydrophobe', 'bool'),
('isaromatic', 'bool'),
('isminus', 'bool'),
('isplus', 'bool'),
('ishalogen', 'bool'),
# secondary structure
('isalpha', 'bool'),
('isbeta', 'bool')
]
a = []
atom_dict = np.empty(self.Mol.GetNumAtoms(), dtype=atom_dtype)
metals = [3,4,11,12,13,19,20,21,22,23,24,25,26,27,28,29,30,31,37,38,39,40,41,42,43,44,45,46,47,48,49,50,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,87,88,89,90,91,
92,93,94,95,96,97,98,99,100,101,102,103]
for i, atom in enumerate(self.atoms):
atomicnum = atom.atomicnum
partialcharge = atom.partialcharge
coords = atom.coords
atomtype = atom.Atom.GetProp("_TriposAtomType") if atom.Atom.HasProp("_TriposAtomType") else atom.Atom.GetSymbol()
if self.protein:
residue = atom.Atom.GetMonomerInfo()
else:
residue = False
# get neighbors, but only for those atoms which realy need them
neighbors = np.zeros(4, dtype=[('coords', 'float32', 3),('atomicnum', 'int8')])
neighbors['coords'].fill(np.nan)
for n, nbr_atom in enumerate(atom.neighbors):
neighbors[n] = (nbr_atom.coords, nbr_atom.atomicnum)
atom_dict[i] = (atom.idx,
coords,
elementtable.GetRvdw(atomicnum),
partialcharge,
atomicnum,
atomtype,
np.clip(atom.Atom.GetHybridization()-1, 0, 3),
neighbors['coords'],
# residue info
residue.GetResidueNumber() if residue else 0,
residue.GetResidueName() if residue else '',
False, # is backbone
# atom properties
False, #IsHbondAcceptor
False, #IsHbondDonor,
False, #IsHbondDonorH,
atomicnum in metals,
atomicnum == 6 and np.in1d(neighbors['atomicnum'], [6,1,0]).all(), #hydrophobe
atom.Atom.GetIsAromatic(),
atomtype in ['O3-', '02-' 'O-'] or atom.formalcharge < 0, # is charged (minus)
atomtype in ['N3+', 'N2+', 'Ng+'] or atom.formalcharge > 0, # is charged (plus)
atomicnum in [9,17,35,53], # is halogen?
False, # alpha
False # beta
)
# Match features and mark them in atom_dict
translate_feats = {
'Donor':'isdonor',
'Acceptor':'isacceptor',
'NegIonizable':'isminus',
'PosIonizable':'isplus',
}
# build residue dictionary
if self.protein:
# for protein finding features per residue is much faster
if self.protein:
for res in self.residues:
for f, field in translate_feats.iteritems():
feats = base_feature_factory.GetFeaturesForMol(res.Residue,includeOnly=f)
atom_dict[field][[res.atommap[idx] for f in feats for idx in f.GetAtomIds()]] = True
res_dict = None
# Protein Residues (alpha helix and beta sheet)
res_dtype = [('id', 'int16'),
('resname', 'a3'),
('N', 'float32', 3),
('CA', 'float32', 3),
('C', 'float32', 3),
('isalpha', 'bool'),
('isbeta', 'bool')
] # N, CA, C
b = []
aa = Chem.MolFromSmarts('[NX3,NX4+][CX4H,CX4H2][CX3](=[OX1])[O,N]') # amino backbone SMARTS
conf = self.Mol.GetConformer()
for path in self.Mol.GetSubstructMatches(aa):
atom_dict['isbackbone'][np.array(path)] = True
residue = self.Mol.GetAtomWithIdx(path[0]).GetMonomerInfo()
b.append((residue.GetResidueNumber(), residue.GetResidueName(), conf.GetAtomPosition(path[0]), conf.GetAtomPosition(path[1]), conf.GetAtomPosition(path[2]), False, False))
res_dict = np.array(b, dtype=res_dtype)
# detect secondary structure by phi and psi angles
first = res_dict[:-1]
second = res_dict[1:]
psi = dihedral(first['N'], first['CA'], first['C'], second['N'])
phi = dihedral(first['C'], second['N'], second['CA'], second['C'])
# mark atoms belonging to alpha and beta
res_mask_alpha = np.where(((phi > -145) & (phi < -35) & (psi > -70) & (psi < 50))) # alpha
res_dict['isalpha'][res_mask_alpha] = True
for i in res_dict[res_mask_alpha]['id']:
atom_dict['isalpha'][atom_dict['resid'] == i] = True
res_mask_beta = np.where(((phi >= -180) & (phi < -40) & (psi <= 180) & (psi > 90)) | ((phi >= -180) & (phi < -70) & (psi <= -165))) # beta
res_dict['isbeta'][res_mask_beta] = True
atom_dict['isbeta'][np.in1d(atom_dict['resid'], res_dict[res_mask_beta]['id'])] = True
else:
# find features for ligands
for f, field in translate_feats.iteritems():
feats = base_feature_factory.GetFeaturesForMol(self.Mol,includeOnly=f)
atom_dict[field][[idx for f in feats for idx in f.GetAtomIds()]] = True
### FIX: remove acidic carbons from isminus group (they are part of smarts)
atom_dict['isminus'][atom_dict['isminus'] & (atom_dict['atomicnum'] == 6)] = False
# Aromatic Rings
r = []
for path in self.sssr:
if self.Mol.GetAtomWithIdx(path[0]).GetIsAromatic():
atom = atom_dict[atom_dict['id'] == path[0]]
coords = atom_dict[np.in1d(atom_dict['id'], path)]['coords']
centroid = coords.mean(axis=0)
# get vector perpendicular to ring
vector = np.cross(coords - np.vstack((coords[1:],coords[:1])), np.vstack((coords[1:],coords[:1])) - np.vstack((coords[2:],coords[:2]))).mean(axis=0) - centroid
r.append((centroid, vector, atom['isalpha'], atom['isbeta']))
ring_dict = np.array(r, dtype=[('centroid', 'float32', 3),('vector', 'float32', 3),('isalpha', 'bool'),('isbeta', 'bool'),])
self._atom_dict = atom_dict
self._atom_dict.setflags(write=False)
self._ring_dict = ring_dict
self._ring_dict.setflags(write=False)
if self.protein:
self._res_dict = res_dict
def _dicts(self):
# Atoms
atom_dtype = [
('id', 'int16'),
# atom info
('coords', 'float16', 3),
('charge', 'float16'),
('atomicnum', 'int8'),
('atomtype', 'a4'),
('hybridization', 'int8'),
(
'neighbors', 'float16', (4, 3)
), # non-H neighbors coordinates for angles (max of 6 neighbors should be enough)
# residue info
('resid', 'int16'),
('resname', 'a3'),
('isbackbone', 'bool'),
# atom properties
('isacceptor', 'bool'),
('isdonor', 'bool'),
('isdonorh', 'bool'),
('ismetal', 'bool'),
('ishydrophobe', 'bool'),
('isaromatic', 'bool'),
('isminus', 'bool'),
('isplus', 'bool'),
('ishalogen', 'bool'),
# secondary structure
('isalpha', 'bool'),
('isbeta', 'bool')
]
a = []
atom_dict = np.empty(self.OBMol.NumAtoms(), dtype=atom_dtype)
metals = [
3, 4, 11, 12, 13, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30,
31, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 55, 56,
57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73,
74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 87, 88, 89, 90, 91, 92, 93,
94, 95, 96, 97, 98, 99, 100, 101, 102, 103
]
for i, atom in enumerate(self.atoms):
atomicnum = atom.atomicnum
# skip non-polar hydrogens for performance
# if atomicnum == 1 and atom.OBAtom.IsNonPolarHydrogen():
# continue
atomtype = typetable.Translate(atom.type) # sybyl atom type
partialcharge = atom.partialcharge
coords = atom.coords
if self.protein:
residue = pybel.Residue(atom.OBAtom.GetResidue())
else:
residue = False
# get neighbors, but only for those atoms which realy need them
neighbors = np.empty(4,
dtype=[('coords', 'float16', 3),
('atomicnum', 'int8')])
neighbors.fill(np.nan)
for n, nbr_atom in enumerate(atom.neighbors):
# concider raising neighbors list to 6, but must do some benchmarks
if n > 3:
break
nbr_atomicnum = nbr_atom.atomicnum
neighbors[n] = (nbr_atom.coords, nbr_atomicnum)
atom_dict[i] = (
atom.idx,
coords,
partialcharge,
atomicnum,
atomtype,
atom.OBAtom.GetHyb(),
neighbors['coords'], #n_coords,
# residue info
residue.idx if residue else 0,
residue.name if residue else '',
residue.OBResidue.GetAtomProperty(atom.OBAtom, 2)
if residue else False, # is backbone
# atom properties
atom.OBAtom.IsHbondAcceptor(),
atom.OBAtom.IsHbondDonor(),
atom.OBAtom.IsHbondDonorH(),
atomicnum in metals,
atomicnum == 6 and not (np.in1d(neighbors['atomicnum'], [
6, 1
])).any(), #hydrophobe #doble negation, since nan gives False
atom.OBAtom.IsAromatic(),
atomtype in ['O3-', '02-'
'O-'], # is charged (minus)
atomtype in ['N3+', 'N2+', 'Ng+'], # is charged (plus)
atomicnum in [9, 17, 35, 53], # is halogen?
False, # alpha
False # beta
)
if self.protein:
# Protein Residues (alpha helix and beta sheet)
res_dtype = [('id', 'int16'), ('resname', 'a3'),
('N', 'float16', 3), ('CA', 'float16', 3),
('C', 'float16', 3), ('isalpha', 'bool'),
('isbeta', 'bool')] # N, CA, C
b = []
for residue in self.residues:
backbone = {}
for atom in residue:
if residue.OBResidue.GetAtomProperty(atom.OBAtom, 1):
if atom.atomicnum == 7:
backbone['N'] = atom.coords
elif atom.atomicnum == 6:
if atom.type == 'C3':
backbone['CA'] = atom.coords
else:
backbone['C'] = atom.coords
if len(backbone.keys()) == 3:
b.append((residue.idx, residue.name, backbone['N'],
backbone['CA'], backbone['C'], False, False))
res_dict = np.array(b, dtype=res_dtype)
# detect secondary structure by phi and psi angles
first = res_dict[:-1]
second = res_dict[1:]
psi = dihedral(first['N'], first['CA'], first['C'], second['N'])
phi = dihedral(first['C'], second['N'], second['CA'], second['C'])
# mark atoms belonging to alpha and beta
res_mask_alpha = np.where(((phi > -145) & (phi < -35) & (psi > -70)
& (psi < 50))) # alpha
res_dict['isalpha'][res_mask_alpha] = True
for i in res_dict[res_mask_alpha]['id']:
atom_dict['isalpha'][atom_dict['resid'] == i] = True
res_mask_beta = np.where(
((phi >= -180) & (phi < -40) & (psi <= 180) & (psi > 90)) |
((phi >= -180) & (phi < -70) & (psi <= -165))) # beta
res_dict['isbeta'][res_mask_beta] = True
atom_dict['isbeta'][np.in1d(atom_dict['resid'],
res_dict[res_mask_beta]['id'])] = True
# Aromatic Rings
r = []
for ring in self.sssr:
if ring.IsAromatic():
path = ring._path
atom = atom_dict[atom_dict['id'] == path[0]]
coords = atom_dict[np.in1d(atom_dict['id'], path)]['coords']
centroid = coords.mean(axis=0)
# get vector perpendicular to ring
vector = np.cross(
coords - np.vstack((coords[1:], coords[:1])),
np.vstack((coords[1:], coords[:1])) - np.vstack(
(coords[2:], coords[:2]))).mean(axis=0) - centroid
r.append((centroid, vector, atom['isalpha'], atom['isbeta']))
ring_dict = np.array(r,
dtype=[
('centroid', 'float16', 3),
('vector', 'float16', 3),
('isalpha', 'bool'),
('isbeta', 'bool'),
])
self._atom_dict = atom_dict
self._ring_dict = ring_dict
if self.protein:
self._res_dict = res_dict
