Hands on with ChEMBL API

Posted on *  •  10 minutes  • 2119 words

ChEMBL is a large-scale bioactivity database that collects information on the interactions between small molecules (such as drugs, compounds, or substances) and their biological targets. It stands for “Chemical Biology Database” and is maintained by the European Molecular Biology Laboratory (EMBL).

Researchers can use ChEMBL to search for specific compounds, investigate their interactions with biological targets (such as enzymes, receptors, or ion channels), and analyze bioactivity data to identify potential drug candidates or understand the mechanisms of action for existing drugs. Additionally, ChEMBL provides tools and APIs (Application Programming Interfaces) for accessing and querying the database programmatically, enabling integration with other bioinformatics and cheminformatics workflows.

import math
from pathlib import Path
from zipfile import ZipFile
from tempfile import TemporaryDirectory

import numpy as np
import pandas as pd
from rdkit.Chem import PandasTools

!pip install chembl-webresource-client

from chembl_webresource_client.new_client import new_client
from tqdm.auto import tqdm

DATA = Path.cwd() / 'data'
DATA.mkdir()

print(DATA)

/Users/sasha/CADD/001_get_bioactivity_data_CHEMBL/data

resource objects for API access.

targets_api = new_client.target
compounds_api = new_client.molecule
bioactivities_api = new_client.activity

Get target data (EGFR kinase: UniProtID : P00533)¶

uniprot_id = "P00533"

# Get target information from ChEMBL for specified values only
targets = targets_api.get(target_components__accession=uniprot_id).only(
    "target_chembl_id", "organism", "pref_name", "target_type"
)
print(f'The type of the targets is "{type(targets)}"')

The type of the targets is "<class 'chembl_webresource_client.query_set.QuerySet'>"

targets = pd.DataFrame.from_records(targets)
targets

Restrict to first entry as our target of iterest

chembl_id = targets.iloc[0].target_chembl_id
print(f"{chembl_id}")

CHEMBL203

Fetch bioactivty data for the target_chembl_id : CHEMBL_203

# fetch the bioactivity data and filter it to only human proteins, IC50, exact measurement, binding data

bioactivities = bioactivities_api.filter(target_chembl_id=chembl_id, type="IC50", relation="=", assay_type="B").only(
    "activity_id",
    "assay_chembl_id",
    "assay_description",
    "assay_type",
    "molecule_chembl_id",
    "type",
    "standard_units",
    "relation",
    "standard_value",
    "target_chembl_id",
    "target_organism",
)

print(f"Length and type of bioactivities object: {len(bioactivities)}, {type(bioactivities)}")

Length and type of bioactivities object: 10420, <class 'chembl_webresource_client.query_set.QuerySet'>

# Whats in here, look at first entry

bioactivities[0]

{'activity_id': 32260,
 'assay_chembl_id': 'CHEMBL674637',
 'assay_description': 'Inhibitory activity towards tyrosine phosphorylation for the epidermal growth factor-receptor kinase',
 'assay_type': 'B',
 'molecule_chembl_id': 'CHEMBL68920',
 'relation': '=',
 'standard_units': 'nM',
 'standard_value': '41.0',
 'target_chembl_id': 'CHEMBL203',
 'target_organism': 'Homo sapiens',
 'type': 'IC50',
 'units': 'uM',
 'value': '0.041'}

# Download into a data frame
bioactivities_df = pd.DataFrame.from_dict(bioactivities)
print(f"DataFrame shape: {bioactivities_df.shape}")
bioactivities_df.head()

DataFrame shape: (10420, 13)

# units has different values
bioactivities_df["units"].unique()

array(['uM', 'nM', 'pM', 'M', "10'3 uM", "10'1 ug/ml", 'ug ml-1',
       "10'-1microM", "10'1 uM", "10'-1 ug/ml", "10'-2 ug/ml", "10'2 uM",
       "10'-3 ug/ml", "10'-2microM", '/uM', "10'-6g/ml", 'mM', 'umol/L',
       'nmol/L', "10'-10M", "10'-7M", 'nmol', '10^-8M', 'µM'],
      dtype=object)

# drop 
bioactivities_df.drop(["units", "value"], axis=1, inplace=True)
bioactivities_df.head()

bioactivities_df.dtypes

activity_id            int64
assay_chembl_id       object
assay_description     object
assay_type            object
molecule_chembl_id    object
relation              object
standard_units        object
standard_value        object
target_chembl_id      object
target_organism       object
type                  object
dtype: object

bioactivities_df = bioactivities_df.astype({"standard_value": "float64"})

Data cleaning

1. Delete missing entries
2. Keep only entries with “standard_unit == nM”
3. Delete duplicates in molecule_chembl_id
4. Rename Columns

# Delete Missing values
bioactivities_df.dropna(axis=0, how="any", inplace=True)
print(f"DataFrame shape: {bioactivities_df.shape}")

DataFrame shape: (10419, 11)

bioactivities_df = bioactivities_df[bioactivities_df["standard_units"] == "nM"]
bioactivities_df.drop_duplicates("molecule_chembl_id", keep="first", inplace=True)
print(f"DataFrame shape: {bioactivities_df.shape}")

DataFrame shape: (6823, 11)

bioactivities_df.reset_index(drop=True, inplace=True)

bioactivities_df.rename(columns={"standard_value": "IC50", "standard_units": "units"}, inplace=True)
bioactivities_df.head()

Fetch compound data (molecule_chembl_id) from ChEMBL

molecule_chembl_id = list(bioactivities_df["molecule_chembl_id"])

compounds_provider = compounds_api.filter(molecule_chembl_id__in = molecule_chembl_id).only("molecule_chembl_id", "molecule_structures")

compounds = list(tqdm(compounds_provider))
compounds_df = pd.DataFrame.from_records(compounds)
print(f"DataFrame shape: {compounds_df.shape}")

100%|██████████| 6823/6823 [06:32<00:00, 17.37it/s]

DataFrame shape: (6823, 2)

compounds_df.head()

Preprocess and filter compound data

1. Remove entries with missing entries
2. Delete duplicate molecules (by molecule_chembl_id)
3. Get molecules with canonical SMILES

compounds_df.dropna(axis=0, how="any", inplace=True)
compounds_df.drop_duplicates("molecule_chembl_id", keep="first", inplace=True)

# Check molecule_structures column
compounds_df.iloc[0].molecule_structures

{'canonical_smiles': 'O=c1oc2c(O)c(O)cc3c(=O)oc4c(O)c(O)cc1c4c23',
 'molfile': '\n     RDKit          2D\n\n 22 25  0  0  0  0  0  0  0  0999 V2000\n   -0.4750   -0.2417    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    0.5750    0.3583    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n   -0.4750   -1.4792    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    0.5750    1.6000    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n   -1.5333    0.3583    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    1.6292   -0.2417    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    0.5750   -2.0875    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n   -0.4750    2.2083    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n   -1.5333    1.6000    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    1.6292   -1.4792    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    1.6292    2.2083    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n   -1.5333   -2.0875    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n   -2.5833   -0.2417    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    2.6792    0.3583    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n   -2.5833   -1.4500    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n    2.6792    1.5625    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0\n   -2.5625    2.2208    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n    2.6625   -2.1042    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n    1.6292    3.4000    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n   -1.5333   -3.2917    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n   -3.6375   -2.0125    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n    3.7375    2.1208    0.0000 O   0  0  0  0  0  0  0  0  0  0  0  0\n  2  1  1  0\n  3  1  1  0\n  4  2  2  0\n  5  1  2  0\n  6  2  1  0\n  7  3  1  0\n  8  9  1  0\n  9  5  1  0\n 10  7  1  0\n 11  4  1  0\n 12  3  2  0\n 13  5  1  0\n 14  6  2  0\n 15 13  2  0\n 16 14  1  0\n 17  9  2  0\n 18 10  2  0\n 19 11  1  0\n 20 12  1  0\n 21 15  1  0\n 22 16  1  0\n 12 15  1  0\n  4  8  1  0\n  6 10  1  0\n 11 16  2  0\nM  END\n> <chembl_id>\nCHEMBL6246\n\n> <chembl_pref_name>\nundefined',
 'standard_inchi': 'InChI=1S/C14H6O8/c15-5-1-3-7-8-4(14(20)22-11(7)9(5)17)2-6(16)10(18)12(8)21-13(3)19/h1-2,15-18H',
 'standard_inchi_key': 'AFSDNFLWKVMVRB-UHFFFAOYSA-N'}

compounds_df.iloc[0].molecule_structures.keys()

dict_keys(['canonical_smiles', 'molfile', 'standard_inchi', 'standard_inchi_key'])

# Keep only canonical_smiles
canonical_smiles = []

for i, compounds in compounds_df.iterrows():
    try:
        canonical_smiles.append(compounds["molecule_structures"]["canonical_smiles"])
    except KeyError:
        canonical_smiles.append(None)

compounds_df["smiles"] = canonical_smiles
compounds_df.shape

(6816, 3)

compounds_df.head()

# Are there missing smiles?
compounds_df[compounds_df["smiles"].isnull()]

compounds_df.drop("molecule_structures", axis=1, inplace=True)
compounds_df.head()

Merge dataframes

bioactivities_df.columns

Index(['activity_id', 'assay_chembl_id', 'assay_description', 'assay_type',
       'molecule_chembl_id', 'relation', 'units', 'IC50', 'target_chembl_id',
       'target_organism', 'type'],
      dtype='object')

compounds_df.columns

Index(['molecule_chembl_id', 'smiles'], dtype='object')

print(f"Bioactivities filtered: {bioactivities_df.shape[0]}")
print(f"Compounds filtered: {compounds_df.shape[0]}")

Bioactivities filtered: 6823
Compounds filtered: 6816

# Merge DataFrames
output_df = pd.merge(
    bioactivities_df[["molecule_chembl_id", "IC50", "units"]],
    compounds_df,
    on="molecule_chembl_id",
)

# Reset row indices
output_df.reset_index(drop=True, inplace=True)

print(f"Dataset with {output_df.shape[0]} entries.")

Dataset with 6816 entries.

output_df.dtypes

molecule_chembl_id     object
IC50                  float64
units                  object
smiles                 object
dtype: object

output_df.head(10)

Convert IC50 to pIC50

def convert_ic50_to_pic50(IC50_value):
    pIC50_value = 9 - math.log10(IC50_value)
    return pIC50_value

# Apply conversion to each row of the compounds DataFrame
output_df["pIC50"] = output_df.apply(lambda x: convert_ic50_to_pic50(x.IC50), axis=1)

output_df.head()

output_df.hist(column="pIC50")

array([[<Axes: title={'center': 'pIC50'}>]], dtype=object)

add a column for RDKit molecule objects

# Add molecule column
PandasTools.AddMoleculeColumnToFrame(output_df, smilesCol="smiles")

# Sort molecules by pIC50
output_df.sort_values(by="pIC50", ascending=False, inplace=True)

# Reset index
output_df.reset_index(drop=True, inplace=True)

#output_df.drop("smiles", axis=1).head(10)

output_df.head(10)

output_df.to_csv(DATA / "EGFR_compounds.csv")