def reactants_graph(tsg):
""" get a graph of the reactants from a transition state graph
"""
frm_bnd_keys = forming_bond_keys(tsg)
ord_dct = dict_.transform_values(bond_orders(tsg), func=round)
gra = set_bond_orders(tsg, ord_dct)
gra = remove_bonds(gra, frm_bnd_keys)
return gra
def resonance_dominant_bond_orders(rgr):
""" resonance-dominant bond orders, by bond
"""
rgr = without_fractional_bonds(rgr)
bnd_keys = list(bond_keys(rgr))
bnd_ords_by_res = [
dict_.values_by_key(bond_orders(dom_rgr), bnd_keys)
for dom_rgr in dominant_resonances(rgr)
]
bnd_ords_lst = list(map(frozenset, zip(*bnd_ords_by_res)))
bnd_dom_res_ords_dct = dict(zip(bnd_keys, bnd_ords_lst))
return bnd_dom_res_ords_dct
def _add_pi_bonds(rgr, bnd_ord_inc_dct):
""" add pi bonds to this graph
"""
bnd_keys = bond_keys(rgr)
assert set(bnd_ord_inc_dct.keys()) <= bnd_keys
bnd_keys = list(bnd_keys)
bnd_ords = dict_.values_by_key(bond_orders(rgr), bnd_keys)
bnd_ord_incs = dict_.values_by_key(bnd_ord_inc_dct, bnd_keys, fill_val=0)
new_bnd_ords = numpy.add(bnd_ords, bnd_ord_incs)
bnd_ord_dct = dict(zip(bnd_keys, new_bnd_ords))
rgr = set_bond_orders(rgr, bnd_ord_dct)
return rgr
def resonance_avg_bond_orders(rgr):
""" resonance-dominant bond orders, by bond
"""
rgr = without_fractional_bonds(rgr)
bnd_keys = list(bond_keys(rgr))
bnd_ords_by_res = [
dict_.values_by_key(bond_orders(dom_rgr), bnd_keys)
for dom_rgr in dominant_resonances(rgr)
]
nres = len(bnd_ords_by_res)
bnd_ords_lst = zip(*bnd_ords_by_res)
avg_bnd_ord_lst = [sum(bnd_ords) / nres for bnd_ords in bnd_ords_lst]
avg_bnd_ord_dct = dict(zip(bnd_keys, avg_bnd_ord_lst))
return avg_bnd_ord_dct
def from_graph(gra):
""" networkx graph object from a molecular graph
"""
nxg = networkx.Graph()
nxg.add_nodes_from(atom_keys(gra))
nxg.add_edges_from(bond_keys(gra))
networkx.set_node_attributes(nxg, atom_symbols(gra), 'symbol')
networkx.set_node_attributes(nxg, atom_implicit_hydrogen_valences(gra),
'implicit_hydrogen_valence')
networkx.set_node_attributes(nxg, atom_stereo_parities(gra),
'stereo_parity')
networkx.set_edge_attributes(nxg, bond_orders(gra), 'order')
networkx.set_edge_attributes(nxg, bond_stereo_parities(gra),
'stereo_parity')
return nxg
def bonds_of_order(gra, mbond=1):
""" Determine the indices of all bonds in a molecule with
the specified bond order.
:param gra: molecular graph
:type gra: molecular graph data structure
:param mbond: bond order of desired bond type
:type mbond: int
"""
# Build resonance graph to get the bond orders
gra = dominant_resonance(gra)
bond_order_dct = bond_orders(gra)
mbond_idxs = tuple()
for bond, order in bond_order_dct.items():
if order == mbond:
bnd1, bnd2 = bond
mbond_idxs += ((bnd1, bnd2), )
return mbond_idxs
def subresonances(rgr):
""" this connected graph and its lower-spin (more pi-bonded) resonances
"""
rgr = without_fractional_bonds(rgr)
# get the bond capacities (room for increasing bond order), filtering out
# the negative ones to avoid complications with hypervalent atoms in TSs
bnd_cap_dct = dict_.by_value(_bond_capacities(rgr), lambda x: x > 0)
ret_rgrs = []
if bnd_cap_dct:
bnd_keys, bnd_caps = zip(*bnd_cap_dct.items())
atm_keys = list(functools.reduce(frozenset.union, bnd_keys))
# Loop over all possible combinations of bond order increments (amounts
# by which to increase the bond order), filtering out combinations that
# exceed the valences of the atoms involved.
# (Note that we are only testing the bonds with available pi electrons,
# so this is compatible with having hypervalent atoms elsewhere in the
# molecule)
bnd_ord_inc_ranges = [range(bnd_cap + 1) for bnd_cap in bnd_caps]
for bnd_ord_incs in itertools.product(*bnd_ord_inc_ranges):
bnd_ord_inc_dct = dict(zip(bnd_keys, bnd_ord_incs))
ret_rgr = _add_pi_bonds(rgr, bnd_ord_inc_dct)
max_bnd_ord = max(bond_orders(ret_rgr).values())
atm_unsat_vlcs = dict_.values_by_key(
atom_unsaturated_valences(ret_rgr), atm_keys)
if not any(atm_unsat_vlc < 0 for atm_unsat_vlc in atm_unsat_vlcs):
if max_bnd_ord < 4:
ret_rgrs.append(ret_rgr)
if not ret_rgrs:
ret_rgrs = (rgr, )
else:
ret_rgrs = tuple(ret_rgrs)
return ret_rgrs
def breaking_bond_keys(tsg):
""" get the forming bonds from a transition state graph
"""
ord_dct = bond_orders(tsg)
brk_bnd_keys = [k for k, o in ord_dct.items() if round(o, 1) == 0.9]
return frozenset(map(frozenset, brk_bnd_keys))
def smiles(gra, stereo=True, local_stereo=False, res_stereo=False):
""" SMILES string from graph
:param gra: molecular graph
:type gra: automol graph data structure
:param stereo: Include stereo?
:type stereo: bool
:param local_stereo: Is the graph using local stereo assignments? That
is, are they based on atom keys rather than canonical keys?
:type local_stereo: bool
:param res_stereo: allow resonant double-bond stereo?
:type res_stereo: bool
:returns: the SMILES string
:rtype: str
"""
assert is_connected(gra), (
"Cannot form connection layer for disconnected graph.")
if not stereo:
gra = without_stereo_parities(gra)
# If not using local stereo assignments, canonicalize the graph first.
# From this point on, the stereo parities can be assumed to correspond to
# the neighboring atom keys.
if not local_stereo:
gra = canonical(gra)
# Convert to implicit graph
gra = implicit(gra)
# Insert hydrogens necessary for bond stereo
gra = _insert_stereo_hydrogens(gra)
# Find a dominant resonance
rgr = dominant_resonance(gra)
# Determine atom symbols
symb_dct = atom_symbols(rgr)
# Determine atom implicit hydrogens
nhyd_dct = atom_implicit_hydrogen_valences(rgr)
# Determine bond orders for this resonance
bnd_ord_dct = bond_orders(rgr)
# Find radical sites for this resonance
rad_atm_keys = radical_atom_keys_from_resonance(rgr)
# Determine neighbors
nkeys_dct = atoms_neighbor_atom_keys(rgr)
# Find stereo parities
atm_par_dct = dict_.filter_by_value(atom_stereo_parities(rgr),
lambda x: x is not None)
bnd_par_dct = dict_.filter_by_value(bond_stereo_parities(rgr),
lambda x: x is not None)
# Remove stereo parities if requested
if not res_stereo:
print('before')
print(bnd_par_dct)
bnd_par_dct = dict_.filter_by_key(bnd_par_dct,
lambda x: bnd_ord_dct[x] == 2)
print('after')
print(bnd_par_dct)
else:
raise NotImplementedError("Not yet implemented!")
def _atom_representation(key, just_seen=None, nkeys=(), closures=()):
symb = ptab.to_symbol(symb_dct[key])
nhyd = nhyd_dct[key]
needs_brackets = key in rad_atm_keys or symb not in ORGANIC_SUBSET
hyd_rep = f'H{nhyd}' if nhyd > 1 else ('H' if nhyd == 1 else '')
par_rep = ''
if key in atm_par_dct:
needs_brackets = True
skeys = [just_seen]
if nhyd:
assert nhyd == 1
skeys.append(-numpy.inf)
if closures:
skeys.extend(closures)
skeys.extend(nkeys)
can_par = atm_par_dct[key]
smi_par = can_par ^ util.is_odd_permutation(skeys, sorted(skeys))
par_rep = '@@' if smi_par else '@'
if needs_brackets:
rep = f'[{symb}{par_rep}{hyd_rep}]'
else:
rep = f'{symb}'
return rep
# Get the pool of stereo bonds for the graph and set up a dictionary for
# storing the ending representation.
ste_bnd_key_pool = list(bnd_par_dct.keys())
drep_dct = {}
def _bond_representation(key, just_seen=None):
key0 = just_seen
key1 = key
# First, handle the bond order
if key0 is None or key1 is None:
rep = ''
else:
bnd_ord = bnd_ord_dct[frozenset({key0, key1})]
if bnd_ord == 1:
rep = ''
elif bnd_ord == 2:
rep = '='
elif bnd_ord == 3:
rep = '#'
else:
raise ValueError("Bond orders greater than 3 not permitted.")
drep = drep_dct[(key0, key1)] if (key0, key1) in drep_dct else ''
bnd_key = next((b for b in ste_bnd_key_pool if key1 in b), None)
if bnd_key is not None:
# We've encountered a new stereo bond, so remove it from the pool
ste_bnd_key_pool.remove(bnd_key)
# Determine the atoms involved
key2, = bnd_key - {key1}
nkey1s = set(nkeys_dct[key1]) - {key2}
nkey2s = set(nkeys_dct[key2]) - {key1}
nmax1 = max(nkey1s)
nmax2 = max(nkey2s)
nkey1 = just_seen if just_seen in nkey1s else nmax1
nkey2 = nmax2
# Determine parity
can_par = bnd_par_dct[bnd_key]
smi_par = can_par if nkey1 == nmax1 else not can_par
# Determine bond directions
drep1 = drep if drep else '/'
if just_seen in nkey1s:
drep = drep1
flip = not smi_par
else:
drep_dct[(key1, nkey1)] = drep1
flip = smi_par
drep2 = _flip_direction(drep1, flip=flip)
drep_dct[(key2, nkey2)] = drep2
rep += drep
# Second, handle directionality (bond stereo)
return rep
# Get the pool of rings for the graph and set up a dictionary for storing
# their tags. As the SMILES is built, each next ring that is encountered
# will be given a tag, removed from the pool, and transferred to the tag
# dictionary.
rng_pool = list(rings_atom_keys(rgr))
rng_tag_dct = {}
def _ring_representation_with_nkeys_and_closures(key, nkeys=()):
nkeys = nkeys.copy()
# Check for new rings in the ring pool. If a new ring is found, create
# a tag, add it to the tags dictionary, and drop it from the rings
# pool.
for new_rng in rng_pool:
if key in new_rng:
# Choose a neighbor key for SMILES ring closure
clos_nkey = sorted(set(new_rng) & set(nkeys))[0]
# Add it to the ring tag dictionary with the current key first
# and the closure key last
tag = max(rng_tag_dct.values(), default=0) + 1
assert tag < 10, (
f"Ring tag exceeds 10 for this graph:\n{string(gra)}")
rng = cycle_ring_atom_key_to_front(new_rng, key, clos_nkey)
rng_tag_dct[rng] = tag
# Remove it from the pool of unseen rings
rng_pool.remove(new_rng)
tags = []
closures = []
for rng, tag in rng_tag_dct.items():
if key == rng[-1]:
nkeys.remove(rng[0])
closures.append(rng[0])
# Handle the special case where the last ring bond has stereo
if (rng[-1], rng[0]) in drep_dct:
drep = drep_dct[(rng[-1], rng[0])]
tags.append(f'{drep}{tag}')
else:
tags.append(f'{tag}')
if key == rng[0]:
nkeys.remove(rng[-1])
closures.append(rng[-1])
tags.append(f'{tag}')
rrep = ''.join(map(str, tags))
return rrep, nkeys, closures
# Determine neighboring keys
nkeys_dct_pool = dict_.transform_values(atoms_neighbor_atom_keys(rgr),
sorted)
def _recurse_smiles(smi, lst, key, just_seen=None):
nkeys = nkeys_dct_pool.pop(key) if key in nkeys_dct_pool else []
# Remove keys just seen from the list of neighbors, to avoid doubling
# back.
if just_seen in nkeys:
nkeys.remove(just_seen)
# Start the SMILES string and connection list. The connection list is
# used for sorting.
rrep, nkeys, closures = _ring_representation_with_nkeys_and_closures(
key, nkeys)
arep = _atom_representation(key, just_seen, nkeys, closures=closures)
brep = _bond_representation(key, just_seen)
smi = f'{brep}{arep}{rrep}'
lst = [key]
# Now, extend the layer/list along the neighboring atoms.
if nkeys:
# Build sub-strings/lists by recursively calling this function.
sub_smis = []
sub_lsts = []
while nkeys:
nkey = nkeys.pop(0)
sub_smi, sub_lst = _recurse_smiles('', [], nkey, just_seen=key)
sub_smis.append(sub_smi)
sub_lsts.append(sub_lst)
# If this is a ring, remove the neighbor on the other side of
# `key` to prevent repetition as we go around the ring.
if sub_lst[-1] == key:
nkeys.remove(sub_lst[-2])
# Now, join the sub-layers and lists together.
# If there is only one neighbor, we joint it as
# {arep1}{brep2}{arep2}...
if len(sub_lsts) == 1:
sub_smi = sub_smis[0]
sub_lst = sub_lsts[0]
# Extend the SMILES string
smi += f'{sub_smi}'
# Extend the list
lst.extend(sub_lst)
# If there are multiple neighbors, we joint them as
# {arep1}({brep2}{arep2}...)({brep3}{arep3}...){brep4}{arep4}...
else:
assert len(sub_lsts) > 1
# Extend the SMILES string
smi += (''.join(map("({:s})".format, sub_smis[:-1])) +
sub_smis[-1])
# Append the lists of neighboring branches.
lst.append(sub_lsts)
return smi, lst
# If there are terminal atoms, start from the first one
atm_keys = atom_keys(rgr)
term_keys = terminal_atom_keys(gra, heavy=False)
start_key = min(term_keys) if term_keys else min(atm_keys)
smi, _ = _recurse_smiles('', [], start_key)
return smi