def test_change_constraints_cache_clear():
alphabet = sf.get_semantic_robust_alphabet()
assert alphabet == sf.get_semantic_robust_alphabet()
assert sf.decoder("[C][#C]") == "C#C"
new_constraints = sf.get_semantic_constraints()
new_constraints["C"] = 1
sf.set_semantic_constraints(new_constraints)
new_alphabet = sf.get_semantic_robust_alphabet()
assert new_alphabet != alphabet
assert sf.decoder("[C][#C]") == "CC"
sf.set_semantic_constraints() # re-set alphabet
def large_alphabet():
alphabet = sf.get_semantic_robust_alphabet()
alphabet.update([
"[#Br]", "[#Branch1]", "[#Branch2]", "[#Branch3]", "[#C@@H1]",
"[#C@@]", "[#C@H1]", "[#C@]", "[#C]", "[#Cl]", "[#F]", "[#H]", "[#I]",
"[#NH1]", "[#N]", "[#O]", "[#P]", "[#Ring1]", "[#Ring2]", "[#Ring3]",
"[#S]", "[/Br]", "[/C@@H1]", "[/C@@]", "[/C@H1]", "[/C@]", "[/C]",
"[/Cl]", "[/F]", "[/H]", "[/I]", "[/NH1]", "[/N]", "[/O]", "[/P]",
"[/S]", "[=Br]", "[=Branch1]", "[=Branch2]", "[=Branch3]", "[=C@@H1]",
"[=C@@]", "[=C@H1]", "[=C@]", "[=C]", "[=Cl]", "[=F]", "[=H]", "[=I]",
"[=NH1]", "[=N]", "[=O]", "[=P]", "[=Ring1]", "[=Ring2]", "[=Ring3]",
"[=S]", "[Br]", "[Branch1]", "[Branch2]", "[Branch3]", "[C@@H1]",
"[C@@]", "[C@H1]", "[C@]", "[C]", "[Cl]", "[F]", "[H]", "[I]", "[NH1]",
"[N]", "[O]", "[P]", "[Ring1]", "[Ring2]", "[Ring3]", "[S]", "[\\Br]",
"[\\C@@H1]", "[\\C@@]", "[\\C@H1]", "[\\C@]", "[\\C]", "[\\Cl]",
"[\\F]", "[\\H]", "[\\I]", "[\\NH1]", "[\\N]", "[\\O]", "[\\P]",
"[\\S]", "[nop]"
])
return list(alphabet)
def hard_alphabet():
"""A challenging alphabet of SELFIES symbols.
"""
alphabet = sf.get_semantic_robust_alphabet()
alphabet.update([
'[#Br]', '[#C@@Hexpl]', '[#C@@expl]', '[#C@Hexpl]', '[#C@expl]',
'[#C]', '[#Cl]', '[#F]', '[#Hexpl]', '[#I]', '[#NHexpl]', '[#N]',
'[#O]', '[#P]', '[#S]', '[/Br]', '[/C@@Hexpl]', '[/C@@expl]',
'[/C@Hexpl]', '[/C@expl]', '[/C]', '[/Cl]', '[/F]', '[/Hexpl]', '[/I]',
'[/NHexpl]', '[/N]', '[/O]', '[/P]', '[/S]', '[=Br]', '[=C@@Hexpl]',
'[=C@@expl]', '[=C@Hexpl]', '[=C@expl]', '[=C]', '[=Cl]', '[=F]',
'[=Hexpl]', '[=I]', '[=NHexpl]', '[=N]', '[=O]', '[=P]', '[=S]',
'[Br]', '[Branch1_1]', '[Branch1_2]', '[Branch1_3]', '[Branch2_1]',
'[Branch2_2]', '[Branch2_3]', '[Branch3_1]', '[Branch3_2]',
'[Branch3_3]', '[C@@Hexpl]', '[C@@expl]', '[C@Hexpl]', '[C@expl]',
'[C]', '[Cl]', '[Expl#Ring1]', '[Expl=Ring1]', '[F]', '[Hexpl]', '[I]',
'[NHexpl]', '[N]', '[O]', '[P]', '[Ring1]', '[Ring2]', '[Ring3]',
'[S]', '[\\Br]', '[\\C@@Hexpl]', '[\\C@@expl]', '[\\C@Hexpl]',
'[\\C@expl]', '[\\C]', '[\\Cl]', '[\\F]', '[\\Hexpl]', '[\\I]',
'[\\NHexpl]', '[\\N]', '[\\O]', '[\\P]', '[\\S]', '[epsilon]', '[nop]'
])
return list(alphabet)
from rdkit import Chem
from selfies import encoder, decoder, get_semantic_robust_alphabet
import re
import random
import logging as logger
ALPHABET = [symbol.strip("[]") for symbol in get_semantic_robust_alphabet()]
# follow these rules for which types of substitution should be allowed
ALLOWED_SUBS = {
"C": ["N", "O", "H"],
"=C": ["=N", "N", "O", "=O", "S"],
"F": ["Cl", "Br", "I", "O", "N", "C", "H"],
"Cl": ["F", "Br", "I", "O", "N", "C", "H"],
"Br": ["Cl", "F", "I", "O", "N", "C", "H"],
"I": ["Cl", "Br", "F", "O", "N", "C", "H"],
"H": ["C", "O", "N", "S", "=C", "=O", "=S"],
"O": ["S", "N", "=O", "=N", "C", "=C", "Cl", "F", "Br", "I", "H"],
"=O": ["=S", "=N", "=C", "O"],
"N": ["O", "C", "H"],
"=N": ["=O", "O", "S", "=C", "C"],
"#N": ["#C"],
"S": ["O", "N", "C", "=O", "=N", "H"],
"=S": ["=O", "=N", "=C", "O"],
"#C": ["#N"]
}
def selfies_substitution(*,
parent_smiles: str,
def mutate_selfie(selfie, max_molecules_len, write_fail_cases=False):
'''Return a mutated selfie string (only one mutation on slefie is performed)
Mutations are done until a valid molecule is obtained
Rules of mutation: With a 33.3% propbabily, either:
1. Add a random SELFIE character in the string
2. Replace a random SELFIE character with another
3. Delete a random character
Parameters:
selfie (string) : SELFIE string to be mutated
max_molecules_len (int) : Mutations of SELFIE string are allowed up to this length
write_fail_cases (bool) : If true, failed mutations are recorded in "selfie_failure_cases.txt"
Returns:
selfie_mutated (string) : Mutated SELFIE string
smiles_canon (string) : canonical smile of mutated SELFIE string
'''
valid = False
fail_counter = 0
chars_selfie = get_selfie_chars(selfie)
while not valid:
fail_counter += 1
alphabet = list(
selfies.get_semantic_robust_alphabet()) # 34 SELFIE characters
choice_ls = [1, 2, 3] # 1=Insert; 2=Replace; 3=Delete
random_choice = np.random.choice(choice_ls, 1)[0]
# Insert a character in a Random Location
if random_choice == 1:
random_index = np.random.randint(len(chars_selfie) + 1)
random_character = np.random.choice(alphabet, size=1)[0]
selfie_mutated_chars = chars_selfie[:random_index] + [
random_character
] + chars_selfie[random_index:]
# Replace a random character
elif random_choice == 2:
random_index = np.random.randint(len(chars_selfie))
random_character = np.random.choice(alphabet, size=1)[0]
if random_index == 0:
selfie_mutated_chars = [random_character
] + chars_selfie[random_index + 1:]
else:
selfie_mutated_chars = chars_selfie[:random_index] + [
random_character
] + chars_selfie[random_index + 1:]
# Delete a random character
elif random_choice == 3:
random_index = np.random.randint(len(chars_selfie))
if random_index == 0:
selfie_mutated_chars = chars_selfie[random_index + 1:]
else:
selfie_mutated_chars = chars_selfie[:
random_index] + chars_selfie[
random_index + 1:]
else:
raise Exception('Invalid Operation trying to be performed')
selfie_mutated = "".join(x for x in selfie_mutated_chars)
sf = "".join(x for x in chars_selfie)
try:
smiles = decoder(selfie_mutated)
mol, smiles_canon, done = sanitize_smiles(smiles)
if len(selfie_mutated_chars
) > max_molecules_len or smiles_canon == "":
done = False
if done:
valid = True
else:
valid = False
except:
valid = False
if fail_counter > 1 and write_fail_cases == True:
f = open("selfie_failure_cases.txt", "a+")
f.write('Tried to mutate SELFIE: ' + str(sf) + ' To Obtain: ' +
str(selfie_mutated) + '\n')
f.close()
return (selfie_mutated, smiles_canon)
c='black',
markersize=7,
linewidth=3) # TARGETS
clb.set_label('Joint Similarity', fontsize=10)
ax.set_xlabel('LogP', fontsize=10)
ax.set_ylabel('QED', fontsize=10)
plt.xlim([-4, 8])
ax.grid(True)
fig.tight_layout()
# plt.savefig('./logP_v_QED_scatter.png', dpi=1000)
plt.show()
alphabet = list(selfies.get_semantic_robust_alphabet()) # 34 SELFIE characters
max_len_random_struct = max([
len(get_selfie_chars(encoder(starting_smile))),
len(get_selfie_chars(encoder(target_smile)))
])
min_len_random_struct = min([
len(get_selfie_chars(encoder(starting_smile))),
len(get_selfie_chars(encoder(target_smile)))
])
num_samples = len(logP_path)
random_selfies = []
for _ in range(num_samples):
selfie = ''
for i in range(random.randint(min_len_random_struct, max_len_random_struct)