def Molecule_volume(mol=Chem.rdchem.Mol(), gv=GaussianVolume()):
EPS = 0.03
N = mol.GetNumAtoms()
for i in range(N):
i += 1
gv.childOverlaps.append([])
gv.gaussians.append(AtomGaussian())
gv.levels.append(N)
gv.volume = 0.0
gv.centroid = np.array([0.0, 0.0, 0.0])
# Stores the parents of gv.gaussians[i] inside parents[i]
parents = [[] for i in range(N)]
# Stores the atom index that have intersection with i_th gaussians inside overlaps[i]
overlaps = [set() for i in range(N)]
atomIndex = 0
vecIndex = N #Used to indicated the initial position of the child gaussian
guassian_weight = 2.828427125
conf = mol.GetConformer()
for atom in mol.GetAtoms():
if atom.GetAtomicNum() == 1: continue
gv.gaussians[atomIndex].centre = np.array(
conf.GetAtomPosition(
atomIndex)) # value chacked, same with mol file
gv.gaussians[atomIndex].alpha = GAlpha(atom.GetAtomicNum())
gv.gaussians[atomIndex].weight = guassian_weight
#radius_VDW = Chem.GetPeriodicTable().GetRvdw(atom.GetAtomicNum())
radius_VDW = ob.GetVdwRad(atom.GetAtomicNum())
'''it looks like the GetRvdw function in rdkit give 1.95 for Carbon,
which is the vdw radius for Br in our paper, here I redefined the value'''
#gv.gaussians[atomIndex].volume = (4.0 * np.pi/3.0) * radius_VDW **3
gv.gaussians[atomIndex].volume = SHoverlap(gv.gaussians[atomIndex],
gv.gaussians[atomIndex])[0]
#checked, give the same value as (np.pi/gv.gaussians[atomIndex].alpha)**1.5 * gv.gaussians[atomIndex].weight
gv.gaussians[atomIndex].n = 1
'''Update volume and centroid of the Molecule'''
gv.volume += gv.gaussians[atomIndex].volume
gv.centroid += gv.gaussians[atomIndex].volume * gv.gaussians[
atomIndex].centre
'''loop over every atom to find the second level overlap'''
for i in range(atomIndex):
ga = atomIntersection(gv.gaussians[i], gv.gaussians[atomIndex])
# Check if overlap is sufficient enough
if ga.volume / (gv.gaussians[i].volume +
gv.gaussians[atomIndex].volume - ga.volume) < EPS:
continue
gv.gaussians.append(ga)
gv.childOverlaps.append([])
#append a empty list in the end to store child of this overlap gaussian
parents.append([i, atomIndex])
overlaps.append(set())
gv.volume -= ga.volume
gv.centroid -= ga.volume * ga.centre
overlaps[i].add(atomIndex)
# store the position of the child (vecIndex) in the root (i)
gv.childOverlaps[i].append(vecIndex)
vecIndex += 1
atomIndex += 1
startLevel = atomIndex
nextLevel = len(gv.gaussians)
gv.levels.append(nextLevel)
LEVEL = 6
for l in range(2, LEVEL):
for i in range(startLevel, nextLevel):
# parents[i] is a pair list e.g.[a1,a2]
a1 = parents[i][0]
a2 = parents[i][1]
# find elements that overlaps with both gaussians(a1 and a2)
overlaps[i] = overlaps[a1] & overlaps[a2]
if len(overlaps[i]) == 0: continue
for elements in overlaps[i]:
# check if there is a wrong index
if elements <= a2: continue
ga = atomIntersection(gv.gaussians[i], gv.gaussians[elements])
if ga.volume / (gv.gaussians[i].volume +
gv.gaussians[elements].volume -
ga.volume) < EPS:
continue
gv.gaussians.append(ga)
#append a empty list in the end to store child of this overlap gaussian
gv.childOverlaps.append([])
parents.append([i, elements])
overlaps.append(set())
if (ga.n % 2
) == 0: # even number overlaps give positive contribution
gv.volume -= ga.volume
gv.centroid -= ga.volume * ga.centre
else: # odd number overlaps give negative contribution
gv.volume += ga.volume
gv.centroid += ga.volume * ga.centre
# store the position of the child (vecIndex) in the root (i)
gv.childOverlaps[i].append(vecIndex)
vecIndex += 1
startLevel = nextLevel
nextLevel = len(gv.gaussians)
gv.levels.append(nextLevel)
overlaps.clear() #!!! why so complacated in C++ code?
parents.clear()
gv.overlap = Molecule_overlap(gv, gv)
return gv
def _dicts(self):
max_neighbors = 6 # max of 6 neighbors should be enough
# Atoms
atom_dtype = [
('id', np.uint32),
# atom info
('coords', np.float32, 3),
('radius', np.float32),
('charge', np.float32),
('atomicnum', np.int8),
('atomtype', 'U5' if PY3 else 'a5'),
('hybridization', np.int8),
('numhs', np.uint8),
('formalcharge', np.int8),
('neighbors_id', np.int16, max_neighbors),
('neighbors', np.float32, (max_neighbors, 3)),
# residue info
('resid', np.int16),
('resnum', np.int16),
('resname', 'U3' if PY3 else '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)
]
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 = Residue(atom.OBAtom.GetResidue())
else:
residue = False
# get neighbors, but only for those atoms which realy need them
neighbors = np.zeros(max_neighbors,
dtype=[('id', np.int16),
('coords', np.float32, 3),
('atomicnum', np.int8)])
neighbors['coords'].fill(np.nan)
for n, nbr_atom in enumerate(atom.neighbors):
if n >= max_neighbors:
warnings.warn(
'Error while parsing molecule "%s" '
'for `atom_dict`. Atom #%i (%s) has %i '
'neighbors (max_neighbors=%i). Additional '
'neighbors are ignored.' %
(self.title, atom.idx0, atomtype, len(
atom.neighbors), max_neighbors), UserWarning)
break
if nbr_atom.atomicnum == 1:
continue
neighbors[n] = (nbr_atom.idx0, nbr_atom.coords,
nbr_atom.atomicnum)
assert i == atom.idx0
atom_dict[i] = (
i,
coords,
ob.GetVdwRad(atomicnum),
partialcharge,
atomicnum,
atomtype,
atom.OBAtom.GetHyb(),
atom.OBAtom.GetTotalDegree() - atom.OBAtom.GetHvyDegree(),
atom.formalcharge,
neighbors['id'],
neighbors['coords'],
# residue info
residue.idx0 if residue else 0,
residue.number if residue else 0,
residue.name if residue else '',
residue.OBResidue.GetAtomProperty(atom.OBAtom, 2)
if residue else False, # is backbone
# atom properties
False, # atom.OBAtom.IsHbondAcceptor(),
False, # atom.OBAtom.IsHbondDonor(),
False, # atom.OBAtom.IsHbondDonorH(),
atomicnum in metals,
atomicnum == 6 and np.in1d(neighbors['atomicnum'],
[6, 1, 0]).all(), # hydrophobe
atom.OBAtom.IsAromatic(),
atom.formalcharge < 0, # is charged (minus)
atom.formalcharge > 0, # is charged (plus)
atomicnum in [9, 17, 35, 53], # is halogen?
False, # alpha
False # beta
)
not_carbon = np.argwhere(
~np.in1d(atom_dict['atomicnum'], [1, 6])).flatten()
# Acceptors
patt = Smarts('[$([O;H1;v2]),'
'$([O;H0;v2;!$(O=N-*),'
'$([O;-;!$(*-N=O)]),'
'$([o;+0])]),'
'$([n;+0;!X3;!$([n;H1](cc)cc),'
'$([$([N;H0]#[C&v4])]),'
'$([N&v3;H0;$(Nc)])]),'
'$([F;$(F-[#6]);!$(FC[F,Cl,Br,I])])]')
matches = np.array(patt.findall(self)).flatten()
if len(matches) > 0:
atom_dict['isacceptor'][np.intersect1d(matches - 1,
not_carbon)] = True
# Donors
patt = Smarts(
'[$([N&!H0&v3,N&!H0&+1&v4,n&H1&+0,$([$([Nv3](-C)(-C)-C)]),'
'$([$(n[n;H1]),'
'$(nc[n;H1])])]),'
# Guanidine can be tautormeic - e.g. Arginine
'$([NX3,NX2]([!O,!S])!@C(!@[NX3,NX2]([!O,!S]))!@[NX3,NX2]([!O,!S])),'
'$([O,S;H1;+0])]')
matches = np.array(patt.findall(self)).flatten()
if len(matches) > 0:
atom_dict['isdonor'][np.intersect1d(matches - 1,
not_carbon)] = True
atom_dict['isdonorh'][[
n.idx0 for idx in np.argwhere(atom_dict['isdonor']).flatten()
for n in self.atoms[int(idx)].neighbors if n.atomicnum == 1
]] = True
# Basic group
patt = Smarts(
'[$([N;H2&+0][$([C,a]);!$([C,a](=O))]),'
'$([N;H1&+0]([$([C,a]);!$([C,a](=O))])[$([C,a]);!$([C,a](=O))]),'
'$([N;H0&+0]([C;!$(C(=O))])([C;!$(C(=O))])[C;!$(C(=O))]),'
'$([N,n;X2;+0])]')
matches = np.array(patt.findall(self)).flatten()
if len(matches) > 0:
atom_dict['isplus'][np.intersect1d(matches - 1, not_carbon)] = True
# Acidic group
patt = Smarts('[CX3](=O)[OX1H0-,OX2H1]')
matches = np.array(patt.findall(self)).flatten()
if len(matches) > 0:
atom_dict['isminus'][np.intersect1d(matches - 1,
not_carbon)] = True
if self.protein:
# Protein Residues (alpha helix and beta sheet)
res_dtype = [('id', np.int16), ('resnum', np.int16),
('resname', 'U3' if PY3 else 'a3'),
('N', np.float32, 3), ('CA', np.float32, 3),
('C', np.float32, 3), ('O', np.float32, 3),
('isalpha', bool), ('isbeta', bool)] # N, CA, C, O
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
elif atom.atomicnum == 8:
backbone['O'] = atom.coords
if len(backbone.keys()) == 4:
b.append((residue.idx0, residue.number, residue.name,
backbone['N'], backbone['CA'], backbone['C'],
backbone['O'], False, False))
res_dict = np.array(b, dtype=res_dtype)
res_dict = detect_secondary_structure(res_dict)
alpha_mask = np.in1d(atom_dict['resid'],
res_dict[res_dict['isalpha']]['id'])
atom_dict['isalpha'][alpha_mask] = True
beta_mask = np.in1d(atom_dict['resid'],
res_dict[res_dict['isbeta']]['id'])
atom_dict['isbeta'][beta_mask] = True
# Aromatic Rings
r = []
for ring in self.sssr:
if ring.IsAromatic():
path = [x - 1 for x in ring._path] # NOTE: mol.sssr is 1-based
atoms = atom_dict[canonize_ring_path(path)]
if len(atoms):
atom = atoms[0]
coords = atoms['coords']
centroid = coords.mean(axis=0)
# get vector perpendicular to ring
ring_vectors = coords - centroid
vector = np.cross(ring_vectors,
np.roll(ring_vectors, shift=-1,
axis=0)).mean(axis=0)
r.append(
(centroid, vector, atom['resid'], atom['resnum'],
atom['resname'], atom['isalpha'], atom['isbeta']))
ring_dict = np.array(r,
dtype=[('centroid', np.float32, 3),
('vector', np.float32, 3),
('resid', np.int16), ('resnum', np.int16),
('resname', 'U3' if PY3 else 'a3'),
('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
self._res_dict.setflags(write=False)
def vdw_radius(atomic_num):
return ob.GetVdwRad(atomic_num)