Don't worry about downloading/extracting ChEMBL or versioning - just use chembl_downloader to write code that knows how to download it and use it automatically.

Install with:

$ pip install chembl-downloader

Full technical documentation can be found on ReadTheDocs. Tutorials can be found in Jupyter notebooks in the notebooks/ directory of the repository.

import chembl_downloader

path = chembl_downloader.download_extract_sqlite(version='28')

After it's been downloaded and extracted once, it's smart and does not need to download again. It gets stored using pystow automatically in the ~/.data/chembl directory.

We'd like to implement something such that it could load directly into SQLite from the archive, but it appears this is a paid feature.

You can modify the previous code slightly by omitting the version keyword argument to automatically find the latest version of ChEMBL:

import chembl_downloader

path = chembl_downloader.download_extract_sqlite()

The version keyword argument is available for all functions in this package (e.g., including connect(), cursor(), and query()), but will be omitted below for brevity.

Inside the archive is a single SQLite database file. Normally, people manually untar this folder then do something with the resulting file. Don't do this, it's not reproducible! Instead, the file can be downloaded and a connection can be opened automatically with:

import chembl_downloader

with chembl_downloader.connect() as conn:
    with conn.cursor() as cursor:
        cursor.execute(...)  # run your query string
        rows = cursor.fetchall()  # get your results

The cursor() function provides a convenient wrapper around this operation:

import chembl_downloader

with chembl_downloader.cursor() as cursor:
    cursor.execute(...)  # run your query string
    rows = cursor.fetchall()  # get your results

The most powerful function is query() which builds on the previous connect() function in combination with pandas.read_sql to make a query and load the results into a pandas DataFrame for any downstream use.

import chembl_downloader

sql = """
SELECT
    MOLECULE_DICTIONARY.chembl_id,
    MOLECULE_DICTIONARY.pref_name
FROM MOLECULE_DICTIONARY
JOIN COMPOUND_STRUCTURES ON MOLECULE_DICTIONARY.molregno == COMPOUND_STRUCTURES.molregno
WHERE molecule_dictionary.pref_name IS NOT NULL
LIMIT 5
"""

df = chembl_downloader.query(sql)
df.to_csv(..., sep='\t', index=False)

Suggestion 1: use pystow to make a reproducible file path that's portable to other people's machines (e.g., it doesn't have your username in the path).

Suggestion 2: RDKit is now pip-installable with pip install rdkit-pypi, which means most users don't have to muck around with complicated conda environments and configurations. One of the powerful but understated tools in RDKit is the rdkit.Chem.PandasTools module.

This example is a bit more fit-for-purpose than the last two. The supplier() function makes sure that the latest SDF dump is downloaded and loads it from the gzip file into a rdkit.Chem.ForwardSDMolSupplier using a context manager to make sure the file doesn't get closed until after parsing is done. Like the previous examples, it can also explicitly take a version.

from rdkit import Chem

import chembl_downloader

with chembl_downloader.supplier() as suppl:
    data = []
    for i, mol in enumerate(suppl):
        if mol is None or mol.GetNumAtoms() > 50:
            continue
        fp = Chem.PatternFingerprint(mol, fpSize=1024, tautomerFingerprints=True)
        smi = Chem.MolToSmiles(mol)
        data.append((smi, fp))

This example was adapted from Greg Landrum's RDKit blog post on generalized substructure search.

This example uses the supplier() method and RDKit to get SMILES strings from molecules in ChEMBL's SDF file. If you want direct access to the RDKit molecule objects, use supplier().

import chembl_downloader

for smiles in chembl_downloader.iterate_smiles():
    print(smiles)

Building on the supplier() function, the get_substructure_library() makes the preparation of a substructure library automated and reproducible. Additionally, it caches the results of the build, which takes on the order of tens of minutes, only has to be done once and future loading from a pickle object takes on the order of seconds.

The implementation was inspired by Greg Landrum's RDKit blog post, Some new features in the SubstructLibrary. The following example shows how it can be used to accomplish some of the first tasks presented in the post:

from rdkit import Chem

import chembl_downloader

library = chembl_downloader.get_substructure_library()
query = Chem.MolFromSmarts('[O,N]=C-c:1:c:c:n:c:c:1')
matches = library.GetMatches(query)

ChEMBL makes a file containing pre-computed 2048 bit radius 2 morgan fingerprints for each molecule available. It can be downloaded using:

import chembl_downloader

path = chembl_downloader.download_fps()

The version and other keyword arguments are also valid for this function.

The following wraps the download_fps function with chemfp's fingerprint loader:

import chembl_downloader

arena = chembl_downloader.chemfp_load_fps()

The version and other keyword arguments are also valid for this function. More information on working with the arena object can be found here.

If you want to store the data elsewhere using pystow (e.g., in pyobo I also keep a copy of this file), you can use the prefix argument.

import chembl_downloader

# It gets downloaded/extracted to 
# ~/.data/pyobo/raw/chembl/29/chembl_29/chembl_29_sqlite/chembl_29.db
path = chembl_downloader.download_extract_sqlite(prefix=['pyobo', 'raw', 'chembl'])

See the pystow documentation on configuring the storage location further.

The prefix keyword argument is available for all functions in this package (e.g., including connect(), cursor(), and query()).

After installing, run the following CLI command to ensure it and send the path to stdout

$ chembl_downloader

Use --test to show two example queries

$ chembl_downloader --test

Please read the contribution guidelines in CONTRIBUTING.md.

If you'd like to contribute, there's a submodule called chembl_downloader.queries where you can add a useful SQL queries along with a description of what it does for easy importing and reuse.

See who's using chembl-downloader.

chembl-downloader is compatible with all versions of ChEMBL. However, some files are not available for all versions. For example, the SQLite version of the database was first added in release 21 (2015-02-12).