Structural Alert Filters¶

Structural filters can be used to triage or flag compounds that contains a specific molecular pattern or featre that can often be unwanted or undesired. Medchem proposes 3 types of structural filters:

• Common Alerts: a list of filters curated and aggregated from ChEMBL and the public litterature. The majority of this list has been curated by Patrick Walters and proposed at https://github.com/PatWalters/rd_filters.
• NIBR Filters: Novartis screening deck originally proposed in Evolution of Novartis’ Small Molecule Screening Deck Design.
• Eli Lilly Demerits Filters: A set of 275 rules used to identify compounds that may interfere with biological assays. Originally proposed in Rules for Identifying Potentially Reactive or Promiscuous Compounds.

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

Common Alerts¶

You can list all the available default filters.

In [ ]:

mc.structural.CommonAlertsFilters.list_default_available_alerts()

mc.structural.CommonAlertsFilters.list_default_available_alerts()

Out[ ]:

	rule_set_name	smarts	catalog_description	rule_set	source
0	Glaxo	55	Glaxo Wellcome Hard filters	1	ChEMBL
1	Dundee	105	University of Dundee NTD Screening Library Fil...	2	ChEMBL
2	BMS	180	Bristol-Myers Squibb HTS Deck filters	3	ChEMBL
3	PAINS	481	PAINS filters	4	ChEMBL
4	SureChEMBL	166	SureChEMBL Non-MedChem Friendly SMARTS	5	ChEMBL
5	MLSMR	116	NIH MLSMR Excluded Functionality filters (MLSMR)	6	ChEMBL
6	Inpharmatica	91	Unwanted fragments derived by Inpharmatica Ltd.	7	ChEMBL
7	LINT	57	Pfizer lint filters (lint)	8	ChEMBL
8	Alarm-NMR	75	Reactive False Positives in Biochemical Screen...	9	Litterature
9	AlphaScreen-Hitters	6	Structural filters for compounds that may be a...	10	Litterature
10	GST-Hitters	34	Structural filters for compounds may prevent G...	11	Litterature
11	HIS-Hitters	19	Structural filters for compounds prevents the ...	12	Litterature
12	LuciferaseInhibitor	3	Structural filters for compounds that may inhi...	13	Litterature
13	DNABinder	78	Structural filters for compounds that may bind...	14	Litterature
14	Chelator	55	Structural filters for compounds that may inhi...	15	Litterature
15	Frequent-Hitter	15	Structural filters for compounds that are freq...	16	Litterature
16	Electrophilic	119	Structural filters for compounds that could ta...	17	Litterature
17	Genotoxic-Carcinogenicity	117	Structural filters for compounds that may caus...	18	Litterature
18	LD50-Oral	20	Structural filters for compounds that may caus...	19	Litterature
19	Non-Genotoxic-Carcinogenicity	22	Structural filters for compounds that may caus...	20	Litterature
20	Reactive-Unstable-Toxic	335	General very reactive/unstable or Toxic compounds	21	Litterature
21	Skin	155	Skin Sensitization filters (irritables)	22	Litterature
22	Toxicophore	154	General Toxicophores	23	Litterature

Create a CommonAlertsFilters object in order to filter a list of molecules.

In [ ]:

alerts = mc.structural.CommonAlertsFilters()

alerts = mc.structural.CommonAlertsFilters()

By default only the "BMS" set is used but you can specify your own set(s):

In [ ]:

alerts2 = mc.structural.CommonAlertsFilters(alerts_set=["LINT", "Toxicophore"])

alerts2 = mc.structural.CommonAlertsFilters(alerts_set=["LINT", "Toxicophore"])

Let's load a few molecules.

In [ ]:

# 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[ ]:

Apply the filters on the list of molecules.

In [ ]:

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

results.head()

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

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

Out[ ]:

	mol	pass_filter	status	reasons
0	<rdkit.Chem.rdchem.Mol object at 0x7f4e549e5690>	True	ok	None
1	<rdkit.Chem.rdchem.Mol object at 0x7f4e53fc5d20>	False	exclude	Polycyclic aromatic hydrocarbon
2	<rdkit.Chem.rdchem.Mol object at 0x7f4e549ccba0>	False	exclude	aniline
3	<rdkit.Chem.rdchem.Mol object at 0x7f4e549ce8f0>	True	ok	None
4	<rdkit.Chem.rdchem.Mol object at 0x7f4e549b9ee0>	True	ok	None

Display the results.

In [ ]:

rows = results.iloc[:8]

mols = rows["mol"].iloc[:8].tolist()
legends = (
    rows[["pass_filter", "reasons"]].apply(lambda x: f"pass_filter={x[0]}\nreasons={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_filter", "reasons"]].apply(lambda x: f"pass_filter={x[0]}\nreasons={x[1]}", axis=1).tolist() ) dm.to_image(mols, legends=legends, mol_size=(300, 250))

Out[ ]:

NIBR Filters¶

Load the NIBR filters.

In [ ]:

nibr_filters = mc.structural.NIBRFilters()

nibr_filters = mc.structural.NIBRFilters()

Let's load a few molecules.

In [ ]:

# 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[ ]:

Apply the filters on the list of molecules.

In [ ]:

results = nibr_filters(
    mols=data["mol"].tolist(),
    n_jobs=-1,
    progress=True,
    progress_leave=True,
    scheduler="threads",
    keep_details=True,
)

results.head()

results = nibr_filters( mols=data["mol"].tolist(), n_jobs=-1, progress=True, progress_leave=True, scheduler="threads", keep_details=True, ) results.head()

NIBR filtering:   0%|          | 0/50 [00:00<?, ?it/s]

Out[ ]:

	mol	reasons	severity	status	n_covalent_motif	special_mol	pass_filter	details
0	<rdkit.Chem.rdchem.Mol object at 0x7f4e517642e0>	ketals _or_acetals_min(1); steroid_non_arom_mi...	10	exclude	0	2	False	{'name': {0: 'ketals _or_acetals_min(1)', 1: '...
1	<rdkit.Chem.rdchem.Mol object at 0x7f4e515b71b0>	polycyclic_systems_14_atoms_min(1)	0	annotations	0	1	True	{'name': {0: 'polycyclic_systems_14_atoms_min(...
2	<rdkit.Chem.rdchem.Mol object at 0x7f4e5176e030>	None	0	ok	0	0	True	NaN
3	<rdkit.Chem.rdchem.Mol object at 0x7f4e5176c3c0>	None	0	ok	0	0	True	NaN
4	<rdkit.Chem.rdchem.Mol object at 0x7f4e515bcb30>	None	0	ok	0	0	True	NaN

In [ ]:

results.columns.tolist()

results.columns.tolist()

Out[ ]:

['mol',
 'reasons',
 'severity',
 'status',
 'n_covalent_motif',
 'special_mol',
 'pass_filter',
 'details']

Display the results.

In [ ]:

rows = results.iloc[:8]

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

Out[ ]:

Eli Lilly Demerits Filters¶

In [ ]:

# 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[ ]:

In [ ]:

from medchem.structural.lilly_demerits import LillyDemeritsFilters

dfilter = LillyDemeritsFilters()

from medchem.structural.lilly_demerits import LillyDemeritsFilters dfilter = LillyDemeritsFilters()

In [ ]:

results = dfilter(
    mols=data["mol"].tolist(),
    n_jobs=-1,
    progress=True,
    progress_leave=True,
    scheduler="threads",
)

results.head()

results = dfilter( mols=data["mol"].tolist(), n_jobs=-1, progress=True, progress_leave=True, scheduler="threads", ) results.head()

Out[ ]:

	smiles	reasons	step	demerit_score	status	pass_filter	mol
0	CC1(C)OC2CC3C4CCC5=CC(=O)C=CC5(C)C4(F)C(O)CC3(...	michael_rejected	2	NaN	exclude	False	<rdkit.Chem.rdchem.Mol object at 0x7f58d6f23530>
1	C1=CC2=C(C=C1)C1=NC=CN=C1C=C2	phenanthrene_het:D60	2	60.0	flag	True	<rdkit.Chem.rdchem.Mol object at 0x7f58d66f9c40>
2	NC1=CC=CC=C1Cl	aniline_h_newd:D50,aniline_h_ewd:D10	2	60.0	flag	True	<rdkit.Chem.rdchem.Mol object at 0x7f5a7da61ee0>
3	O=C(O)CCC(=O)C1=CC=C(C2=CC=CC=C2)C=C1	NaN	2	0.0	ok	True	<rdkit.Chem.rdchem.Mol object at 0x7f58d6da4c10>
4	CC(C)N1C(=O)C2=CC=CC=C2NS1(=O)=O	NaN	2	0.0	ok	True	<rdkit.Chem.rdchem.Mol object at 0x7f58d5a5b610>

---

-- The End :-)