Medchem Rules¶

Medchem rules can be useful to quickly flag or triage compounds that do not meet certain basic molecular properties criteria (such as MW, TSPA, cLogP, etc).

You can find more informations about the rules and their origins within their docstring or in the API documentation.

The below tutorial will introduce you how to apply those rules to a list of molecules.

??? warning: Avoid blindly applying Medchem filters; you may miss valuable compounds or allow toxins for your specific applications.

In [1]:

import datamol as dm
import pandas as pd

import medchem as mc

import datamol as dm import pandas as pd import medchem as mc

Filtering with rules¶

You can list all the available rules.

In [2]:

mc.rules.RuleFilters.list_available_rules()

mc.rules.RuleFilters.list_available_rules()

Out[2]:

	name	rules	description
0	rule_of_five	MW <= 500 & logP <= 5 & HBD <= 5 & HBA <= 10	leadlike;druglike;small molecule;library design
1	rule_of_five_beyond	MW <= 1000 & logP in [-2, 10] & HBD <= 6 & HBA...	leadlike;druglike;small molecule;library design
2	rule_of_four	MW >= 400 & logP >= 4 & RINGS >=4 & HBA >= 4	PPI inhibitor;druglike
3	rule_of_three	MW <= 300 & logP <= 3 & HBA <= 3 & HBD <= 3 & ...	fragment;building block
4	rule_of_three_extended	MW <= 300 & logP in [-3, 3] & HBA <= 6 & HBD <...	fragment;building block
5	rule_of_two	MW <= 200 & logP <= 2 & HBA <= 4 & HBD <= 2	fragment;reagent;building block
6	rule_of_ghose	MW in [160, 480] & logP in [-0.4, 5.6] & Natom...	leadlike;druglike;small molecule;library design
7	rule_of_veber	rotatable bond <= 10 & TPSA < 140	druglike;leadlike;small molecule;oral
8	rule_of_reos	MW in [200, 500] & logP in [-5, 5] & HBA in [0...	druglike;small molecule;library design;HTS
9	rule_of_chemaxon_druglikeness	MW < 400 & logP < 5 & HBA <= 10 & HBD <= 5 & r...	leadlike;druglike;small molecule
10	rule_of_egan	TPSA in [0, 132] & logP in [-1, 6]	druglike;small molecule;admet;absorption;perme...
11	rule_of_pfizer_3_75	not (TPSA < 75 & logP > 3)	druglike;toxicity;invivo;small molecule
12	rule_of_gsk_4_400	MW <= 400 & logP <= 4	druglike;admet;small molecule
13	rule_of_oprea	HBD in [0, 2] & HBA in [2, 9] & ROTBONDS in [2...	druglike;small molecule
14	rule_of_xu	HBD <= 5 & HBA <= 10 & ROTBONDS in [2, 35] & R...	druglike;small molecule;library design
15	rule_of_cns	MW in [135, 582] & logP in [-0.2, 6.1] & TPSA ...	druglike;CNS;BBB;small molecule
16	rule_of_respiratory	MW in [240, 520] & logP in [-2, 4.7] & HBONDS...	druglike;respiratory;small molecule;nasal;inha...
17	rule_of_zinc	MW in [60, 600] & logP < in [-4, 6] & HBD <= 6...	druglike;small molecule;library design;zinc
18	rule_of_leadlike_soft	MW in [150, 400] & logP < in [-3, 4] & HBD <= ...	leadlike;small molecule;library design;admet
19	rule_of_druglike_soft	MW in [100, 600] & logP < in [-3, 6] & HBD <= ...	druglike;small molecule;library design
20	rule_of_generative_design	MW in [200, 600] & logP < in [-3, 6] & HBD <= ...	druglike;small molecule;de novo design;generat...
21	rule_of_generative_design_strict	MW in [200, 600] & logP < in [-3, 6] & HBD <= ...	druglike;small molecule;de novo design;generat...

You can also filter the rules using the tags in their descriptions:

In [3]:

mc.rules.RuleFilters.list_available_rules("building block")

mc.rules.RuleFilters.list_available_rules("building block")

Out[3]:

	name	rules	description
3	rule_of_three	MW <= 300 & logP <= 3 & HBA <= 3 & HBD <= 3 & ...	fragment;building block
4	rule_of_three_extended	MW <= 300 & logP in [-3, 3] & HBA <= 6 & HBD <...	fragment;building block
5	rule_of_two	MW <= 200 & logP <= 2 & HBA <= 4 & HBD <= 2	fragment;reagent;building block

Given a list of rules, you can create a RuleFilters object in order to filter a list of molecules.

In [4]:

# Create the filter object
rfilter = mc.rules.RuleFilters(
    # You can specifiy a rule as a string or as a callable
    rule_list=["rule_of_five", "rule_of_oprea", "rule_of_cns", "rule_of_leadlike_soft"],
    # You can specify a custom list of names
    rule_list_names=["rule_of_five", "rule_of_oprea", "rule_of_cns", "rule_of_leadlike_soft"],
)

# Create the filter object rfilter = mc.rules.RuleFilters( # You can specifiy a rule as a string or as a callable rule_list=["rule_of_five", "rule_of_oprea", "rule_of_cns", "rule_of_leadlike_soft"], # You can specify a custom list of names rule_list_names=["rule_of_five", "rule_of_oprea", "rule_of_cns", "rule_of_leadlike_soft"], )

In [5]:

# Load a dataset
data = dm.data.solubility()
data = data.sample(50, random_state=20)

dm.to_image(data.iloc[:8]["mol"].tolist(), mol_size=(300, 250))

# Load a dataset data = dm.data.solubility() data = data.sample(50, random_state=20) dm.to_image(data.iloc[:8]["mol"].tolist(), mol_size=(300, 250))

Out[5]:

Apply our rule filters on the list of molecules.

In [6]:

results = rfilter(
    mols=data["mol"].tolist(),
    n_jobs=-1,
    progress=True,
    progress_leave=True,
    scheduler="auto",
    keep_props=False,
    fail_if_invalid=True,
)

results.head()

results = rfilter( mols=data["mol"].tolist(), n_jobs=-1, progress=True, progress_leave=True, scheduler="auto", keep_props=False, fail_if_invalid=True, ) results.head()

Filter by rules:   0%|          | 0/50 [00:00<?, ?it/s]

Out[6]:

	mol	pass_all	pass_any	rule_of_five	rule_of_oprea	rule_of_cns	rule_of_leadlike_soft
0	<rdkit.Chem.rdchem.Mol object at 0x15e7597e0>	False	True	True	False	False	False
1	<rdkit.Chem.rdchem.Mol object at 0x15e90d000>	False	True	True	False	True	True
2	<rdkit.Chem.rdchem.Mol object at 0x15e75f530>	False	True	True	False	False	False
3	<rdkit.Chem.rdchem.Mol object at 0x15e75d8c0>	True	True	True	True	True	True
4	<rdkit.Chem.rdchem.Mol object at 0x15e766030>	False	True	True	False	True	True

You will noticde that the columns pass_all and pass_any will indicate whether a molecule has passed either all or at least one of the rules.

Display the results.

In [7]:

rows = results.iloc[:8]

mols = rows["mol"].iloc[:8].tolist()
legends = rows[["pass_all", "pass_any"]].apply(lambda x: f"pass_all={x[0]} - pass_any={x[1]}", axis=1).tolist()

dm.to_image(mols, legends=legends, mol_size=(300, 250))

rows = results.iloc[:8] mols = rows["mol"].iloc[:8].tolist() legends = rows[["pass_all", "pass_any"]].apply(lambda x: f"pass_all={x[0]} - pass_any={x[1]}", axis=1).tolist() dm.to_image(mols, legends=legends, mol_size=(300, 250))

Out[7]:

Low level API¶

You can use the low level API to filter molecules.

List the available rule by their names.

In [8]:

rule_names = mc.rules.RuleFilters.list_available_rules()["name"].tolist()
rule_names

rule_names = mc.rules.RuleFilters.list_available_rules()["name"].tolist() rule_names

Out[8]:

['rule_of_five',
 'rule_of_five_beyond',
 'rule_of_four',
 'rule_of_three',
 'rule_of_three_extended',
 'rule_of_two',
 'rule_of_ghose',
 'rule_of_veber',
 'rule_of_reos',
 'rule_of_chemaxon_druglikeness',
 'rule_of_egan',
 'rule_of_pfizer_3_75',
 'rule_of_gsk_4_400',
 'rule_of_oprea',
 'rule_of_xu',
 'rule_of_cns',
 'rule_of_respiratory',
 'rule_of_zinc',
 'rule_of_leadlike_soft',
 'rule_of_druglike_soft',
 'rule_of_generative_design',
 'rule_of_generative_design_strict']

The rules are available at mc.rules.basic_rules.*. You can retrieve a rule with:

• the direct import mc.rules.basic_rules.rule_of_five.
• by its name: getattr(mc.rules.basic_rules, "rule_of_five")

Let's do it!

In [9]:

# Get some rules
rule_1 = mc.rules.basic_rules.rule_of_five
rule_2 = getattr(mc.rules.basic_rules, "rule_of_chemaxon_druglikeness")

# Get a molecule
mol = dm.to_mol(" CN(C)CCOC(C1=CC=CC=C1)C1=CC=CC=C1")

mol

# Get some rules rule_1 = mc.rules.basic_rules.rule_of_five rule_2 = getattr(mc.rules.basic_rules, "rule_of_chemaxon_druglikeness") # Get a molecule mol = dm.to_mol(" CN(C)CCOC(C1=CC=CC=C1)C1=CC=CC=C1") mol

Out[9]:

In [10]:

# Apply rule #1
rule_1(mol)

# Apply rule #1 rule_1(mol)

Out[10]:

True

In [11]:

# Apply rule #2
rule_2(mol)

# Apply rule #2 rule_2(mol)

Out[11]:

False

---

-- The End :-)