def test_enhance_atoms():
p_table = Get_periodic_table()
#####
mols = dmy.get_rndethane_mols(distance=True)
atoms = GNR.make_atom_df(mols)
structure_dict = GNR.make_struc_dict(atoms)
BCAI.enhance_structure_dict(structure_dict)
###########
BCAI.enhance_atoms(atoms, structure_dict)
for i, idx in enumerate(atoms['atom_index'].values):
molid = atoms['molecule_name'][i]
mol = 0
for ml_fnd in mols:
if ml_fnd.molid == molid:
mol = ml_fnd
atid = atoms['atom_index'][i]
assert p_table.index(atoms['typestr'][i]) == mol.types[atid]
assert np.array_equal(atoms['conn'][i], mol.conn[atid])
assert np.array_equal(atoms['distance'][i], mol.dist[atid])
def flag_to_target(flag):
p_table = Get_periodic_table()
if len(flag) == 4:
if str(flag[0]) == 'J':
length = int(flag[1])
else:
length = int(flag[0])
atype1 = int(p_table.index(str(flag[2])))
atype2 = int(p_table.index(str(flag[3])))
if atype1 >= atype2:
return [length, atype1, atype2]
else:
return [length, atype2, atype1]
elif len(flag) == 3:
atype = int(p_table.index(str(flag[0])))
return [atype]
else:
print('flag, ', flag,
' not recognised, coupling flag format is <nJxy> . . .')
print('flag, ', flag,
' not recognised, chemical shift flag format is <XCS> . . .')
return 0
def test_make_bonds_df():
p_table = Get_periodic_table()
#####
mols = dmy.get_rndethane_mols()
#####
bonds = GNR.make_bonds_df(mols)
assert len(bonds["molecule_name"].unique()) == len(mols)
for idx, bond in enumerate(bonds):
molid = bonds['molecule_name'][idx]
at1 = bonds['atom_index_0'][idx]
at2 = bonds['atom_index_1'][idx]
mol = 0
for ml_fnd in mols:
if ml_fnd.molid == molid:
mol = ml_fnd
assert mol.coupling_len[at1][at2] == int(bonds['type'][idx][0])
assert mol.coupling[at1][at2] == bonds['scalar_coupling_constant'][idx]
def make_optin(prefs, molname, xyz, types, path=''):
# Input:
# prefs: preferences dictionary
# molname: name of molecule
# xyz: xyz coordinates of conformer
# types: type list of conformer (numeric)
# path: path to molecule folder
# Returns: filename for input file
# Get preferences from prefs
charge = prefs['mol']['charge']
multiplicity = prefs['mol']['multiplicity']
functional = prefs['optimisation']['functional']
basis_set = prefs['optimisation']['basisset']
solvent = prefs['optimisation']['solvent']
direct_cmd_line_opt = prefs['optimisation']['custom_cmd_line']
processors = prefs['optimisation']['processors']
# Get periodic table
Periodic_table = Get_periodic_table()
# Define instruction line for ORCA
instr = '! ' + str(functional) + ' ' + str(basis_set) + ' TightSCF OPT miniprint'
# Add parallel option if multiple processors requested
if processors != 1:
instr += ' PAL{0:<d}'.format(processors)
# Add solvent model/solvent if requested
if solvent != 'none':
instr += ' CPCM(' + solvent + ')'
# If direct line input specified then overwrite all of this
if direct_cmd_line_opt:
instr = direct_cmd_line_opt
# Define input file path/name
infile = path.strip() + molname.strip() + '_OPT.in'
# Construct file strings
strings = []
strings.append(instr)
strings.append('')
strings.append("* xyz {0:<1d} {1:<1d}".format(charge, multiplicity))
for i in range(len(xyz)):
str_type = Periodic_table[types[i]]
string = " {0:<2s} {1:>10.5f} {2:>10.5f} {3:>10.5f}".format(str_type, xyz[i][0], xyz[i][1], xyz[i][2])
strings.append(string)
strings.append('*')
strings.append('')
strings.append('%geom')
strings.append(' AddExtraBonds true # switch on/off assigning bonds to atom pairs that are')
strings.append(' # connected by more than <Max_Length> bonds and are less')
strings.append(' # than <MaxDist> Ang. apart (default true)')
strings.append(' AddExtraBonds_MaxLength 10 # cutoff for number of bonds connecting the two')
strings.append(' # atoms (default 10)')
strings.append(' AddExtraBonds_MaxDist 5 # cutoff for distance between two atoms (default 5 Ang.)')
strings.append('end')
# Write file
with open(infile, 'w') as f_handle:
for string in strings:
print(string, file=f_handle)
return infile
def make_nmrin(prefs, molname, xyz, types, path=''):
# Input:
# prefs: preferences dictionary
# molname: name of molecule
# xyz: xyz coordinates of conformer
# types: type list of conformer (numeric)
# path: path to molecule folder
# Returns: input file path/name
# Get values from preferences
charge = prefs['mol']['charge']
multiplicity = prefs['mol']['multiplicity']
functional = prefs['NMR']['functional']
basis_set = prefs['NMR']['basisset']
aux_basis_set = prefs['NMR']['aux_basis_set']
solvent = prefs['NMR']['solvent']
direct_cmd_line_nmr = prefs['NMR']['custom_cmd_line']
processors = prefs['NMR']['processors']
# Get periodic table
Periodic_table = Get_periodic_table()
# Construct instruction line for ORCE
instr = '! ' + str(functional) + ' ' + str(basis_set) + ' ' + str(aux_basis_set) + ' TightSCF miniprint' + ' NMR '
# Add parallel option if multiple processors requested
if processors != 1:
instr += ' PAL{0:<d}'.format(processors)
# Add solvent model/solvent if requested
if solvent != 'none':
instr += ' CPCM(' + solvent + ')'
# If direct line input specified then overwrite all of this
if direct_cmd_line_nmr:
instr = direct_cmd_line_nmr
# Define input file path/name
infile = path.strip() + molname.strip() + '_NMR.in'
# Construct file strings
strings = []
strings.append(instr)
strings.append("")
strings.append("* xyz {0:<1d} {1:<1d}".format(charge, multiplicity))
for i in range(len(xyz)):
str_type = Periodic_table[types[i]]
string = " {0:<2s} {1:>10.6f} {2:>10.6f} {3:>10.6f}".format(str_type, xyz[i][0], xyz[i][1], xyz[i][2])
strings.append(string)
strings.append('*')
strings.append('%eprnmr')
# Needed for the functional we commonly use, ORCA shouted at me
strings.append(" GIAO_2el = GIAO_2el_RIJCOSX")
for type in prefs['NMR']['shift_nuclei']:
strings.append(" Nuclei = all {type:<2s}".format(type=type) + ' { shift }')
for type in prefs['NMR']['spin_nuclei']:
strings.append(" Nuclei = all {type:<2s}".format(type=type) + ' { ssall }')
strings.append('SpinSpinRThresh {0:<f}'.format(prefs['NMR']['spin_thresh']))
strings.append('end')
# Write file
with open(infile, 'w') as f_handle:
for string in strings:
print(string, file=f_handle)
return infile
def print_mol_csv(outname, refs, typerefs, values, labels):
# Input:
# outname: name of output file
# refs: molecular property references (m x k) m = number of properties,
# k = number of atoms in reference + 1,
# ref[y, z] = (molid, atomid1, atomid2, atomid3, . . . )
# typerefs: atom types corresponding to the atom ids in refs (m x k)
# ref[y, z] = (molid, atomtype1, atomtype2, atomtype3, . . . )
# values: molecular properties (n, m, k)
# labels: Labels for molecule sets
# Returns: None
# start empty line array (to print at the end)
lines = []
# get periodic table array
p_table = Get_periodic_table()
# Get number of molecule sets
sets = len(refs[0])
# Get first part of header string (molecule set labels for mol IDs)
idstring = ""
for x in range(len(refs[0])):
idstring = idstring + "{label:<s}MOLID,".format(label=labels[x])
# Get second part of header string (atom references)
refstring = ""
for y in range(1, len(refs[0][0])):
refstring = refstring + "{atom:<s},{type:<s},".format(
atom='Atom',
type='Type',
)
# Get third part of header string (molecule set labels for values)
valstring = ""
for z in range(len(values[0])):
valstring = valstring + "{label:<s}VALUE,".format(label=labels[z])
# Add header string
lines.append(idstring + refstring + valstring)
# Loop through property references
for i in range(len(refs)):
# Get molid string
idstring = ""
for x in range(len(refs[i])):
idstring = idstring + "{id:<s},".format(id=refs[i][x][0])
# Get atomic ref string (atomid, atomtype) x number of atom references
refstring = ""
for y in range(1, len(refs[0][0])):
refstring = refstring + "{atom:<s},{type:<s},".format(
atom=str(refs[i][0][y]), type=p_table[typerefs[i][0][y - 1]])
# Get value string (value1, value2, . . .)
valstring = ""
for z in range(len(values[i])):
valstring = valstring + "{value:<.4f},".format(value=values[i][z])
# Construct line and add to list
lines.append(idstring + refstring + valstring)
# Print all lines to file
with open(outname, 'w') as f:
for line in lines:
print(line, file=f)
def scale_shifts(self, scaling_factors={}):
periodic_table = Get_periodic_table()
for nucleus, factor in scaling_factors.items():
if nucleus in ['basis_set', 'functional']:
continue
for i in range(len(self.shift)):
if periodic_table[self.types[i]] == nucleus:
self.shift[i] = (self.shift[i] - factor[1]) / float(
factor[0])
def labelmaker(i, j, mol):
Periodic_table = Get_periodic_table()
lent = mol.coupling_len[i][j]
label = str(lent) + str('J')
if mol.types[int(i)] >= mol.types[int(j)]:
label = label + str(Periodic_table[mol.types[int(i)]]) + str(Periodic_table[mol.types[int(j)]])
else:
label = label + str(Periodic_table[mol.types[int(j)]]) + str(Periodic_table[mol.types[int(i)]])
return label
def mol_read_type(mol): type_list = [] type_array = np.zeros(len(mol.atoms), dtype=np.int32) Periodic_table = Get_periodic_table() for i in range(len(mol.atoms)): type = int(mol.atoms[i].atomicnum) type_array[i] = type type_list.append(Periodic_table[type]) return type_list, type_array
def test_make_triplets():
p_table = Get_periodic_table()
#####
mols = dmy.get_rndethane_mols(distance=True)
atoms = GNR.make_atom_df(mols)
structure_dict = GNR.make_struc_dict(atoms)
BCAI.enhance_structure_dict(structure_dict)
BCAI.enhance_atoms(atoms, structure_dict)
bonds = GNR.make_bonds_df(mols)
BCAI.enhance_bonds(bonds, structure_dict, flag='3JHH')
############
triplets = BCAI.make_triplets(bonds["molecule_name"].unique(),
structure_dict)
assert len(triplets["molecule_name"].unique()) == len(mols)
count = 0
for mol in mols:
for atom1, type1 in enumerate(mol.types):
for atom2, type2 in enumerate(mol.types):
if atom1 == atom2:
continue
for atom3, type3 in enumerate(mol.types):
if atom3 in [atom1, atom2] or atom3 < atom2:
continue
if mol.conn[atom1][atom2] != 1 or mol.conn[atom1][
atom3] != 1:
continue
row = triplets.loc[(triplets.molecule_name == mol.molid)
& (triplets.atom_index_0 == atom1)
& (triplets.atom_index_1 == atom2)
& (triplets.atom_index_2 == atom3)]
assert len(row.index) == 1
ba = mol.xyz[atom2] - mol.xyz[atom1]
bc = mol.xyz[atom3] - mol.xyz[atom1]
angle = np.sum(
ba * bc) / (np.linalg.norm(ba) * np.linalg.norm(bc))
angle = np.arccos(np.clip(angle, -1.0, 1.0))
assert angle == row.angle.values
count += 1
assert count == len(triplets.index)
def target_to_flag(target):
p_table = Get_periodic_table()
if len(target) == 3:
flag = str(target[0]) + 'J' + str(p_table[target[1]]) + str(
p_table[target[2]])
elif len(target) == 1:
flag = str(target[0]) + 'CS'
else:
print('Error, target ', target, ' not recognised')
return flag
def test_make_struc_dict():
p_table = Get_periodic_table()
mols = dmy.get_rndethane_mols()
atoms = GNR.make_atom_df(mols)
structure_dict = GNR.make_struc_dict(atoms)
assert len(structure_dict.keys()) == len(mols)
for mol in mols:
assert structure_dict[mol.molid]['typesstr'] == [p_table[type] for type in mol.types]
assert np.array_equal(structure_dict[mol.molid]['positions'], mol.xyz)
assert np.array_equal(structure_dict[mol.molid]['conn'], mol.conn)
def test_enhance_bonds():
p_table = Get_periodic_table()
#####
mols = dmy.get_rndethane_mols(distance=True)
atoms = GNR.make_atom_df(mols)
structure_dict = GNR.make_struc_dict(atoms)
BCAI.enhance_structure_dict(structure_dict)
BCAI.enhance_atoms(atoms, structure_dict)
bonds = GNR.make_bonds_df(mols)
############
BCAI.enhance_bonds(bonds, structure_dict, flag='3JHH')
for idx, bond in enumerate(bonds):
molid = bonds['molecule_name'][idx]
at1 = bonds['atom_index_0'][idx]
at2 = bonds['atom_index_1'][idx]
mol = 0
for ml_fnd in mols:
if ml_fnd.molid == molid:
mol = ml_fnd
assert mol.coupling_len[at1][at2] == int(bonds['type'][idx][0])
assert mol.coupling[at1][at2] == bonds['scalar_coupling_constant'][idx]
if bonds['labeled_type'][idx] == '3JHH':
assert bonds['predict'][idx] == 1
else:
assert bonds['predict'][idx] == 0
for mol in mols:
for atom1, type1 in enumerate(mol.types):
for atom2, type2 in enumerate(mol.types):
if atom1 == atom2:
continue
row = bonds.loc[(bonds['molecule_name'] == mol.molid)
& (bonds['atom_index_0'] == atom1)
& (bonds['atom_index_1'] == atom2)]
cpl = row['scalar_coupling_constant'].values
if type1 == 1 and type2 == 1 and mol.coupling_len[atom1][
atom2] == 3:
assert row.predict.values == 1
assert mol.coupling[atom1][atom2] == cpl[0]
def make_atom_df(mols):
p_table = Get_periodic_table()
# construct dataframe as BCAI requires from mols
# atoms has: molecule_name, atom, labeled atom,
molecule_name = [] # molecule name
atom_index = [] # atom index
atom = [] # atom type (letter)
x = [] # x coordinate
y = [] # y coordinate
z = [] # z coordinate
conns = []
mol_order = []
m = -1
for molrf in tqdm(mols, desc='Constructing atom dictionary'):
m += 1
if len(mols) > 2000:
mol = nmrmol(molid=molrf[1])
if molrf[2] == '':
ftype = get_type(molrf[2])
else:
ftype = molrf[2]
mol.read_nmr(molrf[0], ftype)
else:
mol = molrf
mol_order.append(mol.molid)
for t, type in enumerate(mol.types):
molecule_name.append(mol.molid)
atom_index.append(t)
atom.append(p_table[type])
x.append(mol.xyz[t][0])
y.append(mol.xyz[t][1])
z.append(mol.xyz[t][2])
conns.append(mol.conn[t])
atoms = {
'molecule_name': molecule_name,
'atom_index': atom_index,
'atom': atom,
'x': x,
'y': y,
'z': z,
'conn': conns,
}
atoms = pd.DataFrame(atoms)
return atoms
def nmrmol_to_xyz(mol, outname, num=-404):
periodic_table = Get_periodic_table()
with open(outname, 'w') as f:
print(len(mol.types), file=f)
if num == -404:
print(mol.molid, file=f)
else:
string = "{0:<10d}\t{1:<20s}".format(num, mol.molid)
print(string, file=f)
for i in range(len(mol.types)):
string = "{i:<10s}\t{x:<10.6f}\t{y:<10.6f}\t{z:<10.6f}".format(
i=periodic_table[mol.types[i]],
x=mol.xyz[i][0],
y=mol.xyz[i][1],
z=mol.xyz[i][2])
print(string, file=f)
def test_enhance_structure_dict():
p_table = Get_periodic_table()
#####
mols = dmy.get_rndethane_mols(distance=True)
atoms = GNR.make_atom_df(mols)
structure_dict = GNR.make_struc_dict(atoms)
#####
BCAI.enhance_structure_dict(structure_dict)
for mol in mols:
assert structure_dict[mol.molid]['typesstr'] == [
p_table[type] for type in mol.types
]
assert np.array_equal(structure_dict[mol.molid]['positions'], mol.xyz)
assert np.array_equal(structure_dict[mol.molid]['conn'], mol.conn)
assert np.array_equal(structure_dict[mol.molid]['distances'], mol.dist)
def test_add_embedding():
p_table = Get_periodic_table()
#####
mols = dmy.get_rndethane_mols(distance=True)
atoms = GNR.make_atom_df(mols)
structure_dict = GNR.make_struc_dict(atoms)
BCAI.enhance_structure_dict(structure_dict)
BCAI.enhance_atoms(atoms, structure_dict)
bonds = GNR.make_bonds_df(mols)
BCAI.enhance_bonds(bonds, structure_dict, flag='3JHH')
triplets = BCAI.make_triplets(bonds["molecule_name"].unique(),
structure_dict)
#####
embeddings, atoms, bonds, triplets = BCAI.add_embedding(
atoms, bonds, triplets)
def make_atom_df(mols, progress=False):
p_table = Get_periodic_table()
# construct dataframe as BCAI requires from mols
# atoms has: molecule_name, atom, labeled atom,
molecule_name = [] # molecule name
atom_index = [] # atom index
typestr = [] # atom type (string)
typeint = [] # atom type (integer)
x = [] # x coordinate
y = [] # y coordinate
z = [] # z coordinate
conns = []
shifts = []
mol_order = []
m = -1
if progress:
pbar = tqdm(mols, desc='Constructing atom dictionary')
else:
pbar = mols
for molrf in pbar:
m += 1
if len(mols) > 2000:
mol = nmrmol(molid=molrf[1])
if molrf[2] == '':
ftype = get_type(molrf[2])
else:
ftype = molrf[2]
mol.read_nmr(molrf[0], ftype)
else:
mol = molrf
mol_order.append(mol.molid)
for t, type in enumerate(mol.types):
molecule_name.append(mol.molid)
atom_index.append(t)
typestr.append(p_table[type])
typeint.append(type)
x.append(mol.xyz[t][0])
y.append(mol.xyz[t][1])
z.append(mol.xyz[t][2])
conns.append(mol.conn[t])
shifts.append(mol.shift[t])
atoms = {
'molecule_name': molecule_name,
'atom_index': atom_index,
'typestr': typestr,
'typeint': typeint,
'x': x,
'y': y,
'z': z,
'conn': conns,
'shift': shifts
}
atoms = pd.DataFrame(atoms)
return atoms
def nmrmol_to_nmredata(mol, outfile):
# Get periodic table
periodic_table = Get_periodic_table()
# Assume molecule is not chiral
chiral = 0
# Determine number of bonds
bonds = 0
for at1 in range(len(mol.types)):
for at2 in range(at1, len(mol.types)):
if mol.conn[at1][at2] >= 1:
bonds += 1
atoms = len(mol.types)
# Start putting together file lines
lines = []
# print molecule name
if '/' in outfile:
lines.append(outfile.split('/')[-1].split('.')[0])
else:
lines.append(outfile.split('.')[0])
# print file and author
lines.append('auto-ENRICH - 2020 - Will Gerrard')
lines.append('')
# Structure section
string = '{atoms:>3d}{bonds:>3d} 0 0{chiral:>3d} 0 0 0 0 0 1 V2000'.format(
atoms=atoms, bonds=bonds, chiral=chiral)
lines.append(string)
# Print xyz coordinates and types
for i, xyz in enumerate(mol.xyz):
string = '{x:>10.4f}{y:>10.4f}{z:>10.4f} {typechar:>3s} 0 0 0 0 0 0 0 0 0 0 0 0'.format(
x=xyz[0],
y=xyz[1],
z=xyz[2],
typechar=periodic_table[mol.types[i]])
lines.append(string)
# Print bonds and bond types
for at1 in range(len(mol.types)):
for at2 in range(at1, len(mol.types)):
if mol.conn[at1][at2] >= 1:
string = '{at1:>3d}{at2:>3d}{bond:>3d} 0 0 0 0'.format(
at1=at1 + 1, at2=at2 + 1, bond=mol.conn[at1][at2])
lines.append(string)
# Terminate structure section
lines.append('M\tEND'.format())
# assignment section
lines.append('')
lines.append('> <NMREDATA_ASSIGNMENT>')
# Print chemical shifts with variance
for i, shift, type, var in zip(range(len(mol.types)), mol.shift, mol.types,
mol.shift_var):
string = " {atom:<5d}, {shift:<15.8f}, {type:<5d}, {variance:<15.8f}\\".format(
atom=i, shift=shift, type=type, variance=var)
lines.append(string)
lines.append('')
lines.append('> <NMREDATA_J>')
# Print couplings with variance and label
for i in range(len(mol.types)):
for j in range(len(mol.types)):
if i >= j:
continue
if mol.coupling_len[i][j] == 0:
continue
label = labelmaker(i, j, mol)
string = " {a1:<10d}, {a2:<10d}, {coupling:<15.8f}, {label:<10s}, {var:<15.8f}".format(
a1=i,
a2=j,
coupling=mol.coupling[i][j],
label=label,
var=mol.coupling_var[i][j])
lines.append(string)
# Print assembled lines to output file
with open(outfile, 'w') as f:
for line in lines:
print(line, file=f)
def make_bonds_df(mols):
p_table = Get_periodic_table()
# construct dataframe as BCAI requires from mols
# atoms has: molecule_name, atom, labeled atom,
id = [] # number
molecule_name = [] # molecule name
atom_index_0 = [] # atom index for atom 1
atom_index_1 = [] # atom index for atom 2
cpltype = [] # coupling type
coupling = [] # coupling value
r = []
y = []
i = -1
m = -1
for molrf in tqdm(mols, desc='Constructing bond dictionary'):
m += 1
if len(mols) > 2000:
mol = nmrmol(molid=molrf[1])
if molrf[2] == '':
ftype = get_type(molrf[2])
else:
ftype = molrf[2]
mol.read_nmr(molrf[0], ftype)
else:
mol = molrf
moly = []
molr = []
for t, type in enumerate(mol.types):
for t2, type2 in enumerate(mol.types):
if t == t2:
continue
TFM_flag = str(mol.coupling_len[t]
[t2]) + 'J' + p_table[type] + p_table[type2]
#if TFM_flag != flag and flag != 'all':
# continue
i += 1
id.append(i)
molecule_name.append(mol.molid)
atom_index_0.append(t)
atom_index_1.append(t2)
cpltype.append(TFM_flag)
coupling.append(mol.coupling[t][t2])
moly.append(mol.coupling[t][t2])
molr.append([mol.molid, t, t2])
y.append(moly)
r.append(molr)
bonds = {
'id': id,
'molecule_name': molecule_name,
'atom_index_0': atom_index_0,
'atom_index_1': atom_index_1,
'type': cpltype,
'scalar_coupling_constant': coupling
}
bonds = pd.DataFrame(bonds)
return bonds
