ChEMBL Web Services

Web Service Update (May 2017)

A new version of the ChEMBL web services are now available, which expose significantly more data from the under underlying ChEMBL database and introduce new functionality. In addition to the web services that provide access to ChEMBL data, web services that provide access to commonly used cheminformatic methods are also available.

Web Services Documentation

Online interactive documentation is available for both ChEMBL web services:

ChEMBL Data Web Services

The updated ChEMBL web services are not compatible with the previous version of the ChEMBL web services. For this reason we still provide access to the earlier version and the older documentation. We encourage users to migrate any applications using the previous ChEMBL web services to the updated version described on this page.

Getting Started

To best way to get started is to have a look at some example URLs requesting data from the ChEMBL web services. The table below provides a list of examples and a description of the data being returned.

Resources

The following table provides a list of all ChEMBL web service resources currently available.

Supported formats

These are formats currently supported by the API:

from chembl_webresource_client.utils import utils
aspirin = utils.smiles2ctab('O=C(Oc1ccccc1C(=O)O)C')
fingerprints = utils.sdf2fps(aspirin)
      	

Meta Data and Pagination

It is now possible to download all data from a specific ChEMBL web service resource. This is made possible by returning responses from the web services in 'pages', which can be navigated through using a 'page_meta' section. The 'page_meta' section includes information about total number of hits, total number of pages and links to the next and previous pages. An example 'page_meta' section is displayed below:

"page_meta": {

    "limit": 20,
    "next": "/chembl/api/data/activity.json?limit=20&offset=20",
    "offset": 0,
    "previous": null,
    "total_count": 13520737

}

Filtering and Ordering

It is possible to apply search filters to all resource requests using a URL friendly query language. For example, it is possible to return all ChEMBL targets that contain the term 'kinase' in the pref_name attribute with the following URL: https://www.ebi.ac.uk/chembl/api/data/target?pref_name__contains=kinase.

The pattern for applying a filter is as follows:

https://www.ebi.ac.uk/chembl/api/data/[resource]?[field]__[filter_type]=[value]

Examples of other filter type are listed in the table below.

To order the results returned by a particular field the 'order=[field]' argument as added to a request. For example a user can sort targets based on the pref_name using the following URL: https://www.ebi.ac.uk/chembl/api/data/target?order_by=pref_name

The default ordering is in ascending order. To return the results in descending orede place a '-' before the field name: https://www.ebi.ac.uk/chembl/api/data/target?order_by=-pref_name

Note that it is possible combine order_by and filter arguments:https://www.ebi.ac.uk/chembl/api/data/target?pref_name__contains=kinase&order_by=-pref_name

Chemical Searching

The 'Substructure' and 'Similarity' web service resources allow for the chemical content of ChEMBL to be searched. Similar to the other resources, these search based resources except filtering, paging and ordering arguments. These methods accept SMILES, InChI Key and molecule ChEMBL_ID as arguments and in the case of similarity searches an additional identity cut-off is needed. Some example molecule searches are provided in the table below.

Searching with InChI key is only possible for InChI keys found in the ChEMBL database. The system does not try and convert InChI key to a chemical representation.

Molecule Images

The Image resource returns a graphical representation of a ChEMBL molecule. Unlike the other resources it does not except filtering and paging arguments, but does except image specific arguments. These are defined in the table below.

GET, POST and special characters

In GET request all the parameters has to be encoded into URL. Becasue there is a limitation of how long a URL can be it's often more convenient to use POST requests instead. POST parameters are embedded into request body and can be of any size. This is especially importand when retrievieng a long list of entities identified by (random) IDs.

ChEMBL API supports both GET and POST but since POST has a special meaning in REST protocol (CREATE), a special header has to be added to every POST request:

X-HTTP-Method-Override:GET

Another issue is character encoding. SMILES strings often contain characters (such as #, % or \) that have a special meaning in URLs. This is why when using GET, all parameters should be percent-encoded.

One example is a following SMILES string:

[Na+].CO[C@@H](CCC#C\C=C/CCCC(C)CCCCC=C)C(=O)[O-]

Which can be encoded into URL in a following way: https://www.ebi.ac.uk/chembl/api/data/molecule/%5BNa+%5D.CO%5BC@@H%5D(CCC%23C%5CC=C/CCCC(C)CCCCC=C)C(=O)%5BO-%5D

Below is another examle of retrieving a molecule (CHEMBL1628285) that has the longes SMILES string currently stored in ChEMBL. The original SMILES string is:

CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC1OC(CO)C(O)C(O)C1O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC2OC(CO)C(O)C(O)C2O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC3OC(CO)C(O)C(O)C3O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC4OC(CO)C(O)C(O)C4O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC5OC(CO)C(O)C(O)C5O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC6OC(CO)C(O)C(O)C6O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC7OC(CO)C(O)C(O)C7O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC8OC(CO)C(O)C(O)C8O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC9OC(CO)C(O)C(O)C9O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC%10OC(CO)C(O)C(O)C%10O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC%11OC(CO)C(O)C(O)C%11O)C(O)CO.CCCCCCCCCCCCCCCC[NH2+]OC(CO)C(O)C(OC%12OC(CO)C(O)C(O)C%12O)C(O)CO.CCCCCCCCCC(C(=O)NCCc%13ccc(OP(=S)(Oc%14ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%14)N(C)\\N=C\\c%15ccc(Op%16(Oc%17ccc(\\C=N\\N(C)P(=S)(Oc%18ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%18)Oc%19ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%19)cc%17)np(Oc%20ccc(\\C=N\\N(C)P(=S)(Oc%21ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%21)Oc%22ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%22)cc%20)(Oc%23ccc(\\C=N\\N(C)P(=S)(Oc%24ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%24)Oc%25ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%25)cc%23)np(Oc%26ccc(\\C=N\\N(C)P(=S)(Oc%27ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%27)Oc%28ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%28)cc%26)(Oc%29ccc(\\C=N\\N(C)P(=S)(Oc%30ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%30)Oc%31ccc(CCNC(=O)C(CCCCCCCCC)P(=O)(O)[O-])cc%31)cc%29)n%16)cc%15)cc%13)P(=O)(O)[O-]	

After encoding, the URL becomes this

CORS and JSONP

Both techniques are supported

Web Service Client

To help users get started with using the updated ChEMBL web services the existing web service client has also been released. This is written in the Python programming language and is available to install from Python Package Index by typing:

pip install chembl_webresource_client

The following table provides some example use cases of the client.

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
res = molecule.search('viagra')
  		

from chembl_webresource_client.new_client import new_client
target = new_client.target
gene_name = 'GABRB2'
res = target.search(gene_name)
  		

from chembl_webresource_client.new_client import new_client
target = new_client.target
gene_name = 'GABRB2'
res = target.filter(target_synonym__icontains=gene_name)
  		

import csv
from chembl_webresource_client.new_client import new_client

# This will be our resulting structure mapping compound ChEMBL IDs into target uniprot IDs
compounds2targets = dict()

# First, let's just parse the csv file to extract compounds ChEMBL IDs:
with open('compounds_list.csv', 'rb') as csvfile:
    reader = csv.reader(csvfile)
    for row in reader:
        compounds2targets[row[0]] = set()

# OK, we have our source IDs, let's process them in chunks:
chunk_size = 50
keys = compounds2targets.keys()

for i in range(0, len(keys), chunk_size):
    # we jump from compounds to targets through activities:
    activities = new_client.activity.filter(molecule_chembl_id__in=keys[i:i + chunk_size])
    # extracting target ChEMBL IDs from activities:
    for act in activities:
        compounds2targets[act['molecule_chembl_id']].add(act['target_chembl_id'])

# OK, now our dictionary maps from compound ChEMBL IDs into target ChEMBL IDs
# We would like to replace target ChEMBL IDs with uniprot IDs

for key, val in compounds2targets.items():
    # We don't know how many targets are assigned to a given compound so again it's
    # better to process targets in chunks:
    lval = list(val)
    uniprots = set()
    for i in range(0, len(val), chunk_size):
        targets = new_client.target.filter(target_chembl_id__in=lval[i:i + chunk_size])
        uniprots |= set(sum([[comp['accession'] for comp in t['target_components']] for t in targets],[]))
    compounds2targets[key] = uniprots

# Finally write it to the output csv file
with open('compounds_2_targets.csv', 'wb') as csvfile:
    writer = csv.writer(csvfile)
    for key, val in compounds2targets.items():
        writer.writerow([key] + list(val))
  		

import csv
from chembl_webresource_client.new_client import new_client

# This will be our resulting structure mapping compound ChEMBL IDs into target uniprot IDs
compounds2targets = dict()

# First, let's just parse the csv file to extract compounds ChEMBL IDs:
with open('compounds_list.csv', 'rb') as csvfile:
    reader = csv.reader(csvfile)
    for row in reader:
        compounds2targets[row[0]] = set()

# OK, we have our source IDs, let's process them in chunks:
chunk_size = 50
keys = compounds2targets.keys()

for i in range(0, len(keys), chunk_size):
    # we jump from compounds to targets through activities:
    activities = new_client.activity.filter(molecule_chembl_id__in=keys[i:i + chunk_size])
    # extracting target ChEMBL IDs from activities:
    for act in activities:
        compounds2targets[act['molecule_chembl_id']].add(act['target_chembl_id'])

# OK, now our dictionary maps from compound ChEMBL IDs into target ChEMBL IDs
# We would like to replace target ChEMBL IDs with uniprot IDs

for key, val in compounds2targets.items():
    # We don't know how many targets are assigned to a given compound so again it's
    # better to process targets in chunks:
    lval = list(val)
    genes = set()
    for i in range(0, len(val), chunk_size):
        targets = new_client.target.filter(target_chembl_id__in=lval[i:i + chunk_size])
        for target in targets:
            for component in target['target_components']:
                for synonym in component['target_component_synonyms']:
                    if synonym['syn_type'] == "GENE_SYMBOL":
                        genes.add(synonym['component_synonym'])
    compounds2targets[key] = genes

# Finally write it to the output csv file
with open('compounds_2_genes.csv', 'wb') as csvfile:
    writer = csv.writer(csvfile)
    for key, val in compounds2targets.items():
        writer.writerow([key] + list(val))
  		

from chembl_webresource_client.new_client import new_client
similarity = new_client.similarity
res = similarity.filter(smiles="CO[C@@H](CCC#C\C=C/CCCC(C)CCCCC=C)C(=O)[O-]", similarity=85)
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
similarity = new_client.similarity
aspirin_chembl_id = molecule.search('aspirin')[0]['molecule_chembl_id']
res = similarity.filter(chembl_id="CHEMBL25", similarity=70)
  		

from chembl_webresource_client.new_client import new_client
substructure = new_client.substructure
res = substructure.filter(smiles="CN(CCCN)c1cccc2ccccc12")
  		

from chembl_webresource_client.new_client import new_client
substructure = new_client.substructure
substructure.filter(chembl_id="CHEMBL25")
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
m1 = molecule.get('CHEMBL25')
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
m1 = molecule.get('CC(=O)Oc1ccccc1C(=O)O')
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
molecule.get('BSYNRYMUTXBXSQ-UHFFFAOYSA-N')
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
records = molecule.get(['CHEMBL6498', 'CHEMBL6499', 'CHEMBL6505'])
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
records = molecule.get(['CNC(=O)c1ccc(cc1)N(CC#C)Cc2ccc3nc(C)nc(O)c3c2',
      'Cc1cc2SC(C)(C)CC(C)(C)c2cc1\\N=C(/S)\\Nc3ccc(cc3)S(=O)(=O)N',
      'CC(C)C[C@H](NC(=O)[C@@H](NC(=O)[C@H](Cc1c[nH]c2ccccc12)NC(=O)[C@H]3CCCN3C(=O)C(CCCCN)CCCCN)C(C)(C)C)C(=O)O'])
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
records = molecule.get(['XSQLHVPPXBBUPP-UHFFFAOYSA-N', 'JXHVRXRRSSBGPY-UHFFFAOYSA-N', 'TUHYVXGNMOGVMR-GASGPIRDSA-N'])
  		

from chembl_webresource_client.new_client import new_client
activities = new_client.activity
res = activities.filter(molecule_chembl_id="CHEMBL25", pchembl_value__isnull=False)
  		

from chembl_webresource_client.new_client import new_client
activities = new_client.activity
activities.filter(molecule_chembl_id="CHEMBL25", target_chembl_id="CHEMBL612545", pchembl_value__isnull=False)
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
approved_drugs = molecule.filter(max_phase=4)
  		

from chembl_webresource_client.new_client import new_client
drug_indication = new_client.drug_indication
molecules = new_client.molecule
lung_cancer_ind = drug_indication.filter(efo_term__icontains="LUNG CARCINOMA")
lung_cancer_mols = molecules.filter(molecule_chembl_id__in=[x['molecule_chembl_id'] for x in lung_cancer_ind])
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
no_violations = molecule.filter(molecule_properties__num_ro5_violations=0)
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
biotherapeutics = molecule.filter(biotherapeutic__isnull=False)
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
light_molecules = molecule.filter(molecule_properties__mw_freebase__lte=300)
  		

from chembl_webresource_client.new_client import new_client
molecule = new_client.molecule
light_nib_molecules = molecule.filter(molecule_properties__mw_freebase__lte=300).filter(pref_name__iendswith="nib")
  		

from chembl_webresource_client.new_client import new_client
target = new_client.target
activity = new_client.activity
herg = target.search('herg')[0]
herg_activities = activity.filter(target_chembl_id=herg['target_chembl_id']).filter(standard_type="Ki")
  		

from chembl_webresource_client.new_client import new_client
activity = new_client.activity
res = activity.search('"TG-GATES"')
  		

from chembl_webresource_client.new_client import new_client
activity = new_client.activity
res = activity.filter(target_chembl_id='CHEMBL3938', assay_type__iregex='(B|F)')
  		

from chembl_webresource_client.new_client import new_client
assay = new_client.assay
res = assay.search('inhibitor').filter(assay_type='A')
  		

from chembl_webresource_client.new_client import new_client
cell_line = new_client.cell_line
res = cell_line.filter(cellosaurus_id="CVCL_0417")
  		

from chembl_webresource_client.new_client import new_client
drug = new_client.drug
res = drug.filter(first_approval=1976).filter(usan_stem="-azosin")
  		

from chembl_webresource_client.new_client import new_client
tissue = new_client.tissue
res = tissue.filter(bto_id="BTO:0001073")
  		

from chembl_webresource_client.new_client import new_client
tissue = new_client.tissue
res = tissue.filter(caloha_id="TS-0490")
  		

from chembl_webresource_client.new_client import new_client
tissue = new_client.tissue
res = tissue.filter(uberon_id="UBERON:0000173")
  		

from chembl_webresource_client.new_client import new_client
tissue = new_client.tissue
res = tissue.filter(pref_name__istartswith='blood')
  		

from chembl_webresource_client.new_client import new_client
document = new_client.document
res = document.search('cytokine')
  		

from chembl_webresource_client.new_client import new_client
ret = unichem.get('AIN')
  		

from chembl_webresource_client.unichem import unichem_client as unichem
ret = unichem.inchiFromKey('AAOVKJBEBIDNHE-UHFFFAOYSA-N')
  		

from chembl_webresource_client.unichem import unichem_client as unichem
aspirin = utils.smiles2ctab('O=C(Oc1ccccc1C(=O)O)C')
  		

from chembl_webresource_client.utils import utils
aspirin = 'CC(=O)Oc1ccccc1C(=O)O'
im = utils.smiles2image(aspirin)
mol = utils.image2ctab(im)
smiles = utils.ctab2smiles(mol).split()[2]
self.assertEqual(smiles, aspirin)
  		

from chembl_webresource_client.utils import utils
aspirin = utils.smiles2ctab('O=C(Oc1ccccc1C(=O)O)C')
fingerprints = utils.sdf2fps(aspirin)
  		

from chembl_webresource_client.utils import utils
smiles = ["O=C(NCc1cc(OC)c(O)cc1)CCCC/C=C/C(C)C", "CC(C)CCCCCC(=O)NCC1=CC(=C(C=C1)O)OC", "c1(C=O)cc(OC)c(O)cc1"]
mols = [utils.smiles2ctab(smile) for smile in smiles]
sdf = ''.join(mols)
result = utils.mcs(sdf)
  		

from chembl_webresource_client.utils import utils
aspirin = utils.smiles2ctab('O=C(Oc1ccccc1C(=O)O)C')
num_atoms = json.loads(utils.getNumAtoms(aspirin))[0]
mol_wt = json.loads(utils.molWt(aspirin))[0]
log_p = json.loads(utils.logP(aspirin))[0]
tpsa = json.loads(utils.tpsa(aspirin))[0]
descriptors = json.loads(utils.descriptors(aspirin))[0]
  		

from chembl_webresource_client.utils import utils
mol = utils.smiles2ctab("[Na]OC(=O)Cc1ccc(C[NH3+])cc1.c1nnn[n-]1.O")
st = utils.standardise(mol)
  		

Example Queries

The table below provides a list of example searches a user may wish to carry out using the ChEMBL web services. The aim of the list is highlight the type of data that can be retrieved from ChEMBL using the web services. The examples can be adapted, extended and chained togther to build up more complex workflows.

Using the client vs making raw URL queries

There are two alternative methods to accessing the API: through the Python client or via raw URL queries. The advangate of the client is that it hadles pagination and the whole HTTP layer (deciding between GET and POST, character encoding etc.). It also caches results on the client side and provide very convenient interface, similar to Django Query Set.

The advantage of making a raw URL queries is that they can be excuted from any programming language. The following Jupyter (IPython) notebook, compares both methods: https://github.com/chembl/mychembl/blob/master/ipython_notebooks/09_myChEMBL_web_services.ipynb.

Use Cases

The following use cases is provided as an example of how it is possible chain together ChEMBL web service calls to awnser to complicated questions usin ChEMBL data.

Investigating the potency of approved drugs against their efficacy targets

Since ChEMBL includes both mechanism of action information for approved drugs and pharmacology data from published assays, it is interesting to combine this information and investigate the potency of a drug against its efficacy target. This might be done either to confirm or refute the proposed target assignment, or to better understand how the in-vitro potency of a compound might relate to clinical efficacy or ADMET properties. The following example shows how this type of analysis could be carried out using the ChEMBL web services.

1. Use the molecule end point to retrieve a list of approved drugs (max_phase=4):
   
   https://www.ebi.ac.uk/chembl/api/data/molecule?max_phase=4
   
   Using ChEMBL_20, this will retrieve 2795 drugs in total. We will use CHEMBL998 (loratadine) as an example, but the same workflow could be repeated for the others.
   
   
   
2. Use the mechanism end point to retrieve the mechanism of action and target of each drug:
   
   https://www.ebi.ac.uk/chembl/api/data/mechanism?molecule_chembl_id=CHEMBL998
   
   Loratadine is reported to be a histamine H1 receptor antagonist, represented by the ChEMBL target CHEMBL231.
   
   
   
3. Use the assay end point to identify any binding assays (assay_type=B) for the human histamine H1 receptor:
   
   https://www.ebi.ac.uk/chembl/api/data/assay?target_chembl_id=CHEMBL231&relationship_type=D&assay_type=B
   
   Note the relationship_type=D filter restricts the results to assays where we are confident that the human receptor was tested and not an orthologue. A total of 213 assays are identified. We will take CHEMBL1909156 as an example.
   
   
   
4. Combine the results of the above queries to identify any potency measurements for loratadine in these assays:
   
   https://www.ebi.ac.uk/chembl/api/data/activity?molecule_chembl_id=CHEMBL998&assay_chembl_id=CHEMBL1909156
   
   This assay reports an IC50 value of 170nM and a Ki measurement of 20nM for loratadine against the histamine H1 receptor.
   
   This process could be repeated for the other 212 assays by either iterating through them individually or, where a sufficiently small number of assays are returned, using the __in filter on the activity end point to retrieve several assays at once e.g.,
   
   https://www.ebi.ac.uk/chembl/api/data/activity?molecule_chembl_id=CHEMBL998&assay_chembl_id__in=CHEMBL830379,CHEMBL1909156,CHEMBL882906,CHEMBL691450
   
   An additional 2 assays are identified in this way, reporting an IC50 value of 290nM and a Ki value of 414nM. These values could be averaged, or the lowest taken, to give an indication of the average potency of the compound, or the assay conditions could be investigated further to try to identify which assay might be most reliable or informative (the ChEMBL identifier for the document from which the data are extracted is also provided by the activity end point).

More examples

As mentioned above, there is an IPython notebook with example API calls and the corresponsing Python client code https://github.com/chembl/mychembl/blob/master/ipython_notebooks/09_myChEMBL_web_services.ipynb. Additionally there is a comprehensive test suite, covering amost all the functionality offered by the API. It can be found in the Python client library: https://github.com/chembl/chembl_webresource_client/blob/master/chembl_webresource_client/tests.py.

Cheminformatic Utils Web Services

In addition to the described data-focused web services, a utility web service that provides RESTful access to commonly used cheminformatic methods has been released. These web services have been developed under the name 'ChEMBL Beaker' and provide users with access chemical property calculators (e.g. molecular weight), structure format converters (e.g. molfile to SMILES) and a structure standardiser. The methods being exposed by this service are providing RESTful interface to the RDKit cheminformatics library. These services also provide access to methods that extract chemical information from images. The OSRA library is used to expose this functionality. Online documentation is available is available.

Publications

There are a few publications, describing the API from the technical side as well as providing further examples and ideas for using it:

• ChEMBL web services: streamlining access to drug discovery data and utilities
• ChEMBL Beaker: A Lightweight Web Framework Providing Robust and Extensible Cheminformatics Services
• Want Drugs? Use Python

Getting help

If you have any API specific questions please send them to chembl-help@ebi.ac.uk.

Bugs

If you find a bug or any other problem, please consider creating a new issue on GitHub.