def test_scheme_11():
"""SCHEME 11:
Rule 6: Remove Rings of Sizes 3, 5, and 6 First
Rings of sizes 3, 5, and 6 are more frequent and also synthetically more easily accessible than rings of other
sizes. The majority of the commercially available building blocks contain rings of size 3, 5, or 6. If rings of
different sizes occur, they are likely to be built up intentionally to fulfill a dedicated purpose. Often, such
rings are retained throughout a whole series of bioactive compounds. Good examples for this are penicillin,
diazepam, and imipramin together with their related “me-too” analogue compounds.
"""
# 11a
test_smiles = 'C1C2N(C1=O)CCS2'
correct_smiles = canon('O=C1CCN1')
incorrect_smiles = canon('C1CSCN1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags and incorrect_smiles not in frags
# 11b - Epinastine
test_smiles = 'NC1=NCC2N1C1=CC=CC=C1CC1=CC=CC=C21'
correct_smiles = canon('C1=CCNC=CC1')
incorrect_smiles = canon('C1=NCCN1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags and incorrect_smiles not in frags
def test_scheme_19():
""" SCHEME 19:
Whole Result: The whole process is illustrated for Baccatin III
Murcko -> Rule 3 -> Rule 4 -> Rule 4 -> Rule 6
"""
# Baccatin III
test_smiles = 'CC1=C2C(C(=O)C3(C(CC4C(C3C(C(C2(C)C)(CC1O)O)OC(=O)C5=CC=CC=C5)(CO4)OC(=O)C)O)C)OC(=O)C'
original_results = [ # FILTER RULE
canon('O=C(OC1C2CCC=C(CC(=O)C3CCC4OCC4C31)C2)c1ccccc1'), # Murcko
canon('O=C1CC2=CCCC(C2)CC2C1CCC1OCC12'), # Rule 3
canon('O=C1CC2=CCCC(C2)CC2CCCCC12'), # Rule 4
canon('O=C1CCCC2CCC=C(C1)C2'), # Rule 4
canon('O=C1CCCCCCC1') # Rule 6
]
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
for result in original_results:
assert result in frags
assert len(frags) == 5
def test_scheme_9():
""" SCHEME 9:
Rule 4: See scheme 7
"""
# Rhynchophylline
test_smiles = 'CC[C@H]1CN2CC[C@]3(C(=O)Nc4ccccc43)[C@@H]2C[C@@H]1/C(=C\\OC)C(=O)OC'
correct_smiles = canon('O=C1NC=CC12CCNC2')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
def test_scheme_16():
"""SCHEME 16:
Rule 12: Remove Rings First Where the Linker Is Attached to a Ring Hetero-atom at Either End of the Linker
Ring heteroatoms are more easy to functionalise and, therefore, are often functionalised in the later stage
of a chemical library synthesis and thus less characteristic for a chemical scaffold
"""
# Deferasirox
test_smiles = 'O=C1C=CC=C/C1=C1\\N/C(=C2/C=CC=CC2=O)N(c2ccc(C(=O)O)cc2)N1'
correct_smiles = canon('O=C1C=CC=CC1=C1NNC(=C2C=CC=CC2=O)N1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
def test_scheme_8():
""" SCHEME 8:
Rule 4: See scheme 7
"""
# Sophocarpine
test_smiles = 'C1CC2CN3C(CC=CC3=O)C4C2N(C1)CCC4'
correct_smiles = canon('C1CC2CNCC3CCCN(C1)C23')
incorrect_smiles_1 = canon('O=C1C=CCC2C3CCCNC3CCN12')
incorrect_smiles_2 = canon('O=C1C=CCC2CC3NCCCC3CN12')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
assert incorrect_smiles_1 not in frags and incorrect_smiles_2 not in frags
def test_scheme_5():
""" SCHEME 5:
Rule 2: Do Not Remove Rings with ≥ 12 Atoms if There Are Still Smaller Rings To Remove.
If a structure contains a macrocycle, this is considered to be the most characteristic ring system occurring in
the molecule. Therefore, it should be retained. Especially, cyclic peptides may have bicyclic indole side chains
from tryptophane which would be favored by the more general rules
"""
# Seglitide (test smiles)
ts = 'CC(C)C1NC(=O)C(CCCCN)NC(=O)C(Cc2c[nH]c3ccccc23)NC(=O)C(Cc2ccc(O)cc2)NC(=O)C(C)N(C)C(=O)C(Cc2ccccc2)NC1=O'
result_smiles = canon('O=C1CNC(=O)CNC(=O)CNC(=O)CNC(=O)CNC(=O)CN1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(ts))
assert Chem.MolToSmiles(frags[-1]) == result_smiles
def test_scheme_15():
"""SCHEME 15:
Rule 11: For Mixed Aromatic/Non-aromatic Ring Systems, Retain Non-aromatic Rings with Priority
Aromatic systems are extremely frequent, and benzene is the most frequent ring in practically all data sets.
In order to avoid too many compounds to be linked to benzene as the parent scaffold, this rule is introduced
"""
# Sertraline
test_smiles = 'CN[C@H]1CC[C@@H](C2=CC(Cl)=C(Cl)C=C2)C2=CC=CC=C12'
correct_smiles = canon('C1=CCCCC1')
incorrect_smiles = canon('c1ccccc1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags and incorrect_smiles not in frags
def test_scheme_14():
"""SCHEMA 14:
Rule 9: If the Number of Heteroatoms Is Equal, the Priority of Heteroatoms to Retain is N > O > S
This rule is motivated by the important role that N heterocycles play in medicinal chemistry. Sulfur has the
lowest priority, because it is not able to undergo H-bonding.
"""
# Ticlopidine
test_smiles = 'ClC1=CC=CC=C1CN1CCC2=C(C1)C=CS2'
correct_smiles = canon('C1=CCNCC1') # rdkit canonical
incorrect_smiles = canon('c1ccsc1') # rdkit canonical
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags and incorrect_smiles not in frags
def test_scheme_13():
"""SCHEME 13:
Rule 8: Remove Rings with the Least Number of Heteroatoms First
Exocyclic double-bonded heteroatoms (for example, exocyclic carbonyl groups) are not counted as heterocyclic
atoms. For example, in the indole ring, the pyrrol ring is retained instead of the benzene
"""
# Indole
test_smiles = 'C1=C2C(=CC=C1)[N]C=C2'
correct_smiles = canon('C1=C[N]C=C1')
incorrect_smiles = canon('c1ccccc1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags and incorrect_smiles not in frags
def test_scheme_12():
"""SCHEME 12:
Rule 7: A Fully Aromatic Ring System Must Not Be Dissected in a Way That the Resulting System Is Not
Aromatic Any More
The conversion of aromatic in non-aromatic rings is chemically non-intuitive, and would also affect the
geometry of the ring atoms
"""
# Zaleplon
test_smiles = 'O=C(C)N(CC)C1=CC=CC(C2=CC=NC3=C(C=NN23)C#N)=C1'
correct_smiles = canon('c1cn[nH]c1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
def test_scheme_10():
"""SCHEME 10:
Rule 5: Bridged Ring Systems Are Retained with Preference over Spiro Ring Systems
Under certain circumstances, ring systems containing ring fusions as well as bridged rings can be dissected to
produce a spiro ring or alternatively a bridged ring. Both solutions would have the same |Δ| value.
Therefore, in the cases where the two remaining sub-scaffolds have the same value for |Δ|, the ring system with
a positive signed value of Δ is to be retained.
"""
# Cafestol
test_smiles = 'OC[C@@]5(O)C[C@@]31C[C@@H]5CC[C@H]1[C@]4(C)CCc2occc2[C@H]4CC3'
correct_smiles = canon('C1CC2CCC(C1)C2')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
def test_scheme_6():
"""SCHEME 6:
Rule 3: Choose the Parent Scaffold Having the Smallest Number of Acyclic Linker Bonds
This leads to the removal of linked rings before removing fused rings. Rings linked by longer chains are removed
first. Linkers are usually the most likely point of a retrosynthetic disconnection. In the synthesis of
combinatorial libraries, the variable side chains are often attached to a cyclic core by some linking reaction
creating an acyclic linker. Whenever different cyclic side chains are used, their pruning at an early stage leads
to the preservation of the common core of the library. Therefore, it is intuitive to dissect scaffolds at acyclic
linkers. Also, this helps in retaining preferentially more rigid scaffolds which are more likely to have a
unique interaction pattern.
"""
# Flucloxacillin murcko scaffold
test_smiles = 'O=C(NC1C(=O)N2CCSC12)c1conc1-c1ccccc1'
result_smiles = canon('O=C(NC1C(=O)N2CCSC12)c1cnoc1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
assert Chem.MolToSmiles(frags[1]) == result_smiles
def test_scheme_4():
"""SCHEME 4:
Rule 1: Remove Heterocycles of Size 3 First
As an exception to the general rules, the fusion bond connecting the three-membered ring with other rings is
converted into a double bond. This rule is intended to deal with epoxides and aziridines. This rule treats
such systems as functional groups which are removed beforehand, rather than as rings. This reflects the situation
that epoxides are usually generated by the oxidation of a double bond, and also many natural products exist often
in forms with and without epoxidized double bonds
"""
# Epothilone A
test_smiles = 'CC1CCCC2C(O2)CC(OC(=O)CC(C(C(=O)C(C1O)C)(C)C)O)C(=CC3=CSC(=N3)C)C'
# Epothilone C
result_smiles = canon('O=C1CCCCCCC=CCC(C=Cc2cscn2)OC(=O)CCC1')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
assert frags[1].GetRingInfo().NumRings() == 2
frags = [Chem.MolToSmiles(x) for x in frags]
assert result_smiles in frags
def test_scheme_18():
""" SCHEME 18:
Whole result:
The whole process is illustrated on a set of four diazepinenones, one of the best known classes of anxiolytics:
(diazepam, bromazepam, zolazepam, and clotiazepam)
It can be seen that the molecular frameworks of these four drugs are different despite the fact that they
are usually regarded as belonging to the same class of compounds. In all four cases, the linked ring is
removed first according to rule 3. This already leads to the grouping of diazepam and bromazepam into the
same scaffold class, whereas the other two drugs are still in their distinct classes. After the removal of
the five- or six-membered aromatic ring attached to the diazepinenone ring system according to rule 6,
the seven membered diazepinenone ring remaining is equal for all four molecules.
"""
d = 'CN1C(=O)CN=C(C2=C1C=CC(=C2)Cl)C3=CC=CC=C3' # Diazepam
b = 'C1C(=O)NC2=C(C=C(C=C2)Br)C(=N1)C3=CC=CC=N3' # Bromazepam
z = 'CC1=NN(C2=C1C(=NCC(=O)N2C)C3=CC=CC=C3F)C' # Zolazepam
c = 'CCC1=CC2=C(S1)N(C(=O)CN=C2C3=CC=CC=C3Cl)C' # Clotiazepam
original_results = [ # MOLECULES | # HIERARCHY
canon('O=C1CN=C(c2ccccc2)c2ccccc2N1'), # Diazepam (murcko) | (3)
canon('O=C1CN=Cc2ccccc2N1'), # Diazepam + Bromazepam | (2)
canon('O=C1CN=C(c2ccccn2)c2ccccc2N1'), # Bromazepam (murcko) | (3)
canon(
'O=C1CN=C(c2ccccc2)c2cn[nH]c2N1'), # Zolazepam (murcko) | (3)
canon('O=C1CN=Cc2cn[nH]c2N1'), # Zolazepam | (2)
canon('O=C1CN=C(c2ccccc2)c2ccsc2N1'), # Clotiazepam (murcko) | (3)
canon('O=C1CN=Cc2ccsc2N1'), # Clotiazepam | (2)
canon('O=C1CN=CC=CN1'), # ALL | (1)
# < TOTAL: 8 > #
]
molecules = [Chem.MolFromSmiles(x) for x in [d, b, z, c]]
frags = list(chain(*[tree_frags_from_mol(x) for x in molecules]))
frags = {Chem.MolToSmiles(x) for x in frags}
for result in original_results:
assert result in frags
assert len(frags) == 8
def test_scheme_7():
"""SCHEME 7:
Rule 4: Retain Bridged Rings, Spiro Rings, and Nonlinear Ring Fusion Patterns with Preference
These patterns are unusual structural features occurring less frequently than normally fused rings.
They have non-planar, characteristic molecular shapes, which distinguishes them from the majority of the more
planar organic molecules. In most ring systems, we have a linear fusion with no atoms in common to more than
two rings. This is, for example, the case in steroids. In such cases, the number of bonds being a member in more
than one ring nrrb is equal to the number of rings nR − 1. The more bridges or nonlinear ring fusions there are,
the higher the number of nrrb is. On the other hand, nrrb decreases if there are spiro connected ring systems,
because the spiro connections lead to no bond in common to both rings. Therefore, we remove that ring with
preference where the remaining scaffold has the highest value for |Δ| = |nrrb − (nR − 1)|.
"""
# Pentazocine
test_smiles = 'C[C@H]1[C@H]2Cc3ccc(O)cc3[C@]1(C)CCN2CC=C(C)C'
correct_smiles = canon('C1=CC2CCNC(C1)C2')
incorrect_smiles = canon('c1ccc2c(c1)CCCC2')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
assert incorrect_smiles not in frags
def test_scheme_17():
"""SCHEME 17:
Rule 13: Tie-breaking Rule
Remaining ties are solved by choosing from several possible remaining sub-scaffolds that one, whose rdkit canonical
SMILES, has the lower rank in alphabetical order. Although the nature of this tie-breaking rule is arbitrary,
the use of this rule in the classification does not mean that it will lead to a completely arbitrary overall
class assignment.
This result is different to the original publication as rdkit's canonicalisation algorithm is different to
MolInspiration's algorithm used in the original
"""
# Ormeloxifene
test_smiles = 'CC1([C@@H]([C@H](c2ccc(cc2O1)OC)c3ccc(cc3)OCCN4CCCC4)c5ccccc5)C'
hierarchy_3_1 = canon('c1ccc(C2COc3ccccc3C2)cc1')
hierarchy_3_2 = canon('c1ccc(C2CCOc3ccccc32)cc1')
correct_smiles = sorted([hierarchy_3_1, hierarchy_3_2])[0]
hierarchy_2_correct = canon('c1ccc2c(c1)CCCO2')
frags = tree_frags_from_mol(Chem.MolFromSmiles(test_smiles))
frags = [Chem.MolToSmiles(x) for x in frags]
assert correct_smiles in frags
assert hierarchy_2_correct in frags