def _is_compatible(sgr):
atm_ste_par_dct = _assigned(atom_stereo_parities(sgr))
bnd_ste_par_dct = _assigned(bond_stereo_parities(sgr))
_compat_atm_assgns = (set(known_atm_ste_par_dct.items()) <= set(
atm_ste_par_dct.items()))
_compat_bnd_assgns = (set(known_bnd_ste_par_dct.items()) <= set(
bnd_ste_par_dct.items()))
return _compat_atm_assgns and _compat_bnd_assgns
def compatible_reverse_stereomers(ste_tsg):
""" Given a TS graph with stereo assignments, expand all possible reverse
graphs compatble with the forward graph.
:param ste_tsg: The TS graph, with stereo assignments.
:returns: All possible reverse TS graphs.
"""
frm_bnd_keys = forming_bond_keys(ste_tsg)
brk_bnd_keys = breaking_bond_keys(ste_tsg)
_, des_ste_atm_keys = nonconserved_atom_stereo_keys(ste_tsg)
_, des_ste_bnd_keys = nonconserved_bond_stereo_keys(ste_tsg)
cons_atm_keys = sorted(atom_stereo_keys(ste_tsg) - des_ste_atm_keys)
cons_bnd_keys = sorted(bond_stereo_keys(ste_tsg) - des_ste_bnd_keys)
# 1. Determine index-based stereo assignments for conserved stereo centers
idx_tsg = to_index_based_stereo(ste_tsg)
cons_idx_atm_pars = dict_.values_by_key(atom_stereo_parities(idx_tsg),
cons_atm_keys)
cons_idx_bnd_pars = dict_.values_by_key(bond_stereo_parities(idx_tsg),
cons_bnd_keys)
# 2. Determine all possible index-based stereo assignments for the reverse
# reaction.
prds_gra = without_stereo_parities(products_graph(ste_tsg))
prds_sgrs = _stereomers(prds_gra)
prds_idx_sgrs = list(map(_to_index_based_stereo, prds_sgrs))
rev_idx_tsgs_pool = [
graph(p, brk_bnd_keys, frm_bnd_keys) for p in prds_idx_sgrs
]
# 3. Find possibilities which match the assignments for the conserved
# stereo centers.
rev_idx_tsgs = []
for rev_idx_tsg in rev_idx_tsgs_pool:
rev_cons_idx_atm_pars = dict_.values_by_key(
atom_stereo_parities(rev_idx_tsg), cons_atm_keys)
rev_cons_idx_bnd_pars = dict_.values_by_key(
bond_stereo_parities(rev_idx_tsg), cons_bnd_keys)
if (rev_cons_idx_atm_pars == cons_idx_atm_pars
and rev_cons_idx_bnd_pars == cons_idx_bnd_pars):
rev_idx_tsgs.append(rev_idx_tsg)
# 4. Convert the matching reverse graphs back from index-based stereo
# assignments to absolute stereo assignments.
rev_ste_tsgs = list(map(from_index_based_stereo, rev_idx_tsgs))
return rev_ste_tsgs
def to_index_based_stereo(sgr):
""" Convert a graph to index-based stereo assignments, where parities are
defined relative to the ordering of indices rather than the absolute stereo
priority.
:param sgr: a graph with absolute stereo assignments
:returns: a graph with index-based stereo assignments
"""
assert sgr == explicit(sgr), (
f"Not an explicit graph:\n{string(sgr, one_indexed=False)}")
abs_srt_keys_dct = atoms_stereo_sorted_neighbor_atom_keys(sgr)
atm_ste_keys = atom_stereo_keys(sgr)
bnd_ste_keys = bond_stereo_keys(sgr)
abs_atm_ste_par_dct = atom_stereo_parities(sgr)
abs_bnd_ste_par_dct = bond_stereo_parities(sgr)
idx_atm_ste_par_dct = {}
idx_bnd_ste_par_dct = {}
# Determine index-based stereo assignments for atoms
for atm_key in atm_ste_keys:
abs_srt_keys = abs_srt_keys_dct[atm_key]
idx_srt_keys = sorted(abs_srt_keys)
if util.is_even_permutation(idx_srt_keys, abs_srt_keys):
idx_atm_ste_par_dct[atm_key] = abs_atm_ste_par_dct[atm_key]
else:
idx_atm_ste_par_dct[atm_key] = not abs_atm_ste_par_dct[atm_key]
# Determine index-based stereo assignments for bonds
for bnd_key in bnd_ste_keys:
atm1_key, atm2_key = sorted(bnd_key)
atm1_abs_srt_keys = abs_srt_keys_dct[atm1_key]
atm2_abs_srt_keys = abs_srt_keys_dct[atm2_key]
atm1_idx_srt_keys = sorted(atm1_abs_srt_keys)
atm2_idx_srt_keys = sorted(atm2_abs_srt_keys)
if not ((atm1_idx_srt_keys[0] != atm1_abs_srt_keys[0]) ^
(atm2_idx_srt_keys[0] != atm2_abs_srt_keys[0])):
idx_bnd_ste_par_dct[bnd_key] = abs_bnd_ste_par_dct[bnd_key]
else:
idx_bnd_ste_par_dct[bnd_key] = not abs_bnd_ste_par_dct[bnd_key]
sgr = set_atom_stereo_parities(sgr, idx_atm_ste_par_dct)
sgr = set_bond_stereo_parities(sgr, idx_bnd_ste_par_dct)
return sgr
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 _atom_stereo_layer(gra):
""" AMChI atom stereo (t, m) layer from graph
S = counterclockwise = '@' = '-' = False
R = clockwise = '@@' = '+' = True
:param gra: molecular graph
:type gra: automol graph data structure
:rtype: (str, str)
"""
# Generate the atom stereo layer strings
atm_keys = sorted(atom_stereo_keys(gra))
atm_par_dct = atom_stereo_parities(gra)
atm_pars = list(map(atm_par_dct.__getitem__, atm_keys))
atm_sgns = ['+' if p else '-' for p in atm_pars]
atm_strs = [f'{i+1}{s}' for i, s in zip(atm_keys, atm_sgns)]
atm_ste_lyr = ','.join(atm_strs)
return atm_ste_lyr
def substereomers(gra):
""" all stereomers compatible with this graph's assignments
"""
_assigned = functools.partial(dict_.filter_by_value,
func=lambda x: x is not None)
known_atm_ste_par_dct = _assigned(atom_stereo_parities(gra))
known_bnd_ste_par_dct = _assigned(bond_stereo_parities(gra))
def _is_compatible(sgr):
atm_ste_par_dct = _assigned(atom_stereo_parities(sgr))
bnd_ste_par_dct = _assigned(bond_stereo_parities(sgr))
_compat_atm_assgns = (set(known_atm_ste_par_dct.items()) <= set(
atm_ste_par_dct.items()))
_compat_bnd_assgns = (set(known_bnd_ste_par_dct.items()) <= set(
bnd_ste_par_dct.items()))
return _compat_atm_assgns and _compat_bnd_assgns
sgrs = tuple(filter(_is_compatible, stereomers(gra)))
return sgrs
def _stereo_corrected_geometry(sgr, geo, geo_idx_dct):
""" correct the stereo parities of a geometry
(works iterately to handle cases of higher-order stereo)
"""
assert sgr == explicit(sgr)
gra = without_stereo_parities(sgr)
if has_stereo(sgr):
full_atm_ste_par_dct = atom_stereo_parities(sgr)
full_bnd_ste_par_dct = bond_stereo_parities(sgr)
atm_keys = set()
bnd_keys = set()
last_gra = None
while last_gra != gra:
last_gra = gra
atm_keys.update(stereogenic_atom_keys(gra))
bnd_keys.update(stereogenic_bond_keys(gra))
atm_ste_par_dct = {
atm_key: full_atm_ste_par_dct[atm_key]
for atm_key in atm_keys
}
bnd_ste_par_dct = {
bnd_key: full_bnd_ste_par_dct[bnd_key]
for bnd_key in bnd_keys
}
geo, gra = _atom_stereo_corrected_geometry(gra, atm_ste_par_dct,
geo, geo_idx_dct)
geo, gra = _bond_stereo_corrected_geometry(gra, bnd_ste_par_dct,
geo, geo_idx_dct)
return geo
def from_index_based_stereo(sgr):
""" Convert a graph from index-based stereo assignments back to absolute
stereo assignments, where parities are independent of atom ordering.
:param sgr: a graph with index-based stereo assignments
:returns: a graph with absolute stereo assignments
"""
assert sgr == explicit(sgr), (
f"Not an explicit graph:\n{string(sgr, one_indexed=False)}")
gra = without_stereo_parities(sgr)
if has_stereo(sgr):
atm_keys_pool = atom_stereo_keys(sgr)
bnd_keys_pool = bond_stereo_keys(sgr)
idx_atm_ste_par_dct = atom_stereo_parities(sgr)
idx_bnd_ste_par_dct = bond_stereo_parities(sgr)
atm_ngb_keys_dct = atoms_neighbor_atom_keys(sgr)
atm_keys = set()
bnd_keys = set()
last_gra = None
# Do the assignments iteratively to handle higher-order stereo
while last_gra != gra:
last_gra = gra
abs_atm_ste_par_dct = {}
abs_bnd_ste_par_dct = {}
atm_keys.update(stereogenic_atom_keys(gra) & atm_keys_pool)
bnd_keys.update(stereogenic_bond_keys(gra) & bnd_keys_pool)
# Determine absolute stereo assignments for atoms
for atm_key in atm_keys:
abs_srt_keys = atom_stereo_sorted_neighbor_atom_keys(
gra, atm_key, atm_ngb_keys_dct[atm_key])
idx_srt_keys = sorted(abs_srt_keys)
if util.is_even_permutation(idx_srt_keys, abs_srt_keys):
abs_atm_ste_par_dct[atm_key] = (
idx_atm_ste_par_dct[atm_key])
else:
abs_atm_ste_par_dct[atm_key] = (
not idx_atm_ste_par_dct[atm_key])
# Determine absolute stereo assignments for bonds
for bnd_key in bnd_keys:
atm1_key, atm2_key = sorted(bnd_key)
atm1_abs_srt_keys = atom_stereo_sorted_neighbor_atom_keys(
gra, atm1_key, atm_ngb_keys_dct[atm1_key] - bnd_key)
atm2_abs_srt_keys = atom_stereo_sorted_neighbor_atom_keys(
gra, atm2_key, atm_ngb_keys_dct[atm2_key] - bnd_key)
atm1_idx_srt_keys = sorted(atm1_abs_srt_keys)
atm2_idx_srt_keys = sorted(atm2_abs_srt_keys)
if not ((atm1_idx_srt_keys[0] != atm1_abs_srt_keys[0]) ^
(atm2_idx_srt_keys[0] != atm2_abs_srt_keys[0])):
abs_bnd_ste_par_dct[bnd_key] = (
idx_bnd_ste_par_dct[bnd_key])
else:
abs_bnd_ste_par_dct[bnd_key] = (
not idx_bnd_ste_par_dct[bnd_key])
gra = set_atom_stereo_parities(gra, abs_atm_ste_par_dct)
gra = set_bond_stereo_parities(gra, abs_bnd_ste_par_dct)
atm_ste_keys = atom_stereo_keys(gra)
bnd_ste_keys = bond_stereo_keys(gra)
assert atm_ste_keys == atm_keys_pool, (
"Index-based to absolute stereo conversion failed:\n"
f"{str(atm_ste_keys)} != {str(atm_keys_pool)}")
assert bnd_ste_keys == bnd_keys_pool, (
"Index-based to absolute stereo conversion failed:\n"
f"{str(bnd_ste_keys)} != {str(bnd_keys_pool)}")
return gra
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