Annual Scientific Report 2015

ChEMBL
Drug discovery is more costly than ever, and innovation in efficacy and safety
remains a significant challenge. Changes in the pharmaceutical industry over the
past decade have led to an increase in drug-discovery activities in organisations
that typically have access neither to large databases of legacy bioactivity data nor
the experienced staff needed to manage them. Our team develops and manages
ChEMBL, EMBL-EBI’s database of quantitative small-molecule bioactivity
data focused in the area of drug discovery; SureChEMBL, a patent resource
containing chemical structures extracted from patents on a daily basis; and
UniChem, a resource to link chemical structures across databases, both internal
and external to EMBL-EBI.
ChEMBL contains data on curated chemical structures,
bioactivity values and their relationship to biological
targets and phenotypic assays. SureChEMBL combines
full patent text and automatically data-mined chemical
structures, significantly extending the speed and scope
of public data available to drug-discovery researchers.
The combination of structure–activity relationship
(SAR) data from the scientific literature, deposited
data from neglected disease high-throughout screens
and now the patent literature all make ChEMBL an
important and enabling resource for scientists working
in pharmaceutical R&D.
Our research interests centre on data mining the
ChEMBL database for applications relevant to
translational drug discovery, including aspects of genetic
variability, drug safety and neglected diseases.
Major achievements
In 2015 there was a major change in the ChEMBL Group
when John Overington, who had been Team Leader
since the database was taken on by EMBL-EBI in 2008,
left to join the London-based biotech company Stratified
Medical. Since April 2015 Anne Hersey has been Acting
Team Leader.
We continued to expand the data coverage of ChEMBL
to include drug-metabolism and pharmacokinetic
(DMPK) data, and undertook extensive target and
disease annotations on approved drugs and clinical
candidates. We also developed methods to enhance and
streamline the curation of data and significantly updated
our Web Services as a flexible way for users to access
ChEMBL data.
We further refined the SureChEMBL patent annotation
pipeline to improve its robustness and provided and
provided new methods to access the annotations.
The number of databases indexed in UniChem has
increased to 27. We put in place a process to update the
resource automatically every week.
During 2015 ChEMBL data content continued to
expand, with the number of compounds reaching
1.7 million and the number of bioactivities nearly 14
million. Access to the full ChEMBL data continues to
be freely available in a wide variety of technical formats
including a web interface, data downloads, web services
and Semantic Web technologies. During the year there
were approximately 15,000 unique visitors per month
to the web interface on average. There were substantial
increases in the extraction of data from the scientific
literature; in particular, we extracted data on drug
metabolism and disposition and integrated it into the
database. ChEMBL Web Services were significantly
expanded and re-implemented to expose more data
types and provide new functionality. In addition, we
added cheminformatics Web Services based on RDkit
that allow users to perform more complex queries and to
combine data and chemistry-aware queries.
UniChem grew to contain links to over 100 million
chemical structures from 27 source databases. For
example, the UniChem web services are used on the
ChEMBL web interface to provide dynamic links to
other resources via the matching of the InChI/InChI
Key. We fully automated the mechanism of updating and
registering compounds in UniChem and since the start
of 2015 weekly updates have been provided via the web
interface, web services and as downloadable files.
At the end of 2015 the number of novel chemical entities
annotated in SureChEMBL stood at approximately
17 million, growing at a rate of around 80,000 novel
chemicals per month from roughly 50,000 new patents.
Previously, the patent data in SureChEMBL was
available only via a web interface. In 2015, in response to
user demand, we increased options for users to access
the data. We now provide a quarterly download of files
111
2015 EMBL-EBI Annual Scientific Report