How the InChI identifier is used to underpin our online chemistry

databases at RSC

Antony Williams, Valery Tkachenko

and Ken Karapetyan

ACS San Francisco

August 2014

What can I say that I haven’t said?

What can I say that I haven’t said?

What can I say that I haven’t said?YouTube InChIKey Collision Movie

What can I say that I haven’t said?

InChI is for machines but do have a human aspect…

Many Names, One Structure

Structure Identifiers

OPSIN (chemical name to structure) http://opsin.ch.cam.ac.uk/

• InChI support systems…

InChI mapping helps a lot!

• We wanted to map together chemical data on the web

• We knew that chemical name mapping was difficult but dictionaries were useful

• It is InChI that became the foundation technology for our database…

• We accepted all the limitations of InChI• We lived with the “Useful but not ideal”• And so….

• ~32 million chemicals and growing

• Data sourced from >500 different sources

• Crowd sourced curation and annotation

• Ongoing deposition of data from our journals and our collaborators

• Structure centric hub for web-searching

• …and a really big dictionary!!!

ChemSpider

So where can we travel???

InChI String Search via GoogleSo give me InChIKeys…

And where can we travel???

And where can we travel???

And where can we travel???

And where can we travel???

NEW15th

Edition

*The name THE MERCK INDEX is owned by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Whitehouse Station, N.J., U.S.A., and is licensed to The Royal Society of Chemistry for use in the U.S.A. and Canada.

Where else is RSC using InChIs

Text Mining

The N-(β-hydroxyethyl)-N-methyl-N'-(2-trifluoromethyl-1,3,4-thiadiazol-5-yl)urea prepared in Example 6 , thionyl chloride ( 5 ml ) and benzene ( 50 ml ) were charged into a glass reaction vessel equipped with a mechanical stirrer , thermometer and reflux condenser .

The reaction mixture was heated at reflux with stirring , for a period of about one-half hour .

After this time the benzene and unreacted thionyl chloride were stripped from the reaction mixture under reduced pressure to yield the desired product N-(β-chloroethyl)-N-methyl-N'-(2-trifluoromethyl-1,3,4-thiaidazol-5-yl)urea as a solid residue

Text Mining

The N-(β-hydroxyethyl)-N-methyl-N'-(2-trifluoromethyl-1,3,4-thiadiazol-5-yl)urea prepared in Example 6 , thionyl chloride ( 5 ml ) and benzene ( 50 ml ) were charged into a glass reaction vessel equipped with a mechanical stirrer , thermometer and reflux condenser .

The reaction mixture was heated at reflux with stirring , for a period of about one-half hour .

After this time the benzene and unreacted thionyl chloride were stripped from the reaction mixture under reduced pressure to yield the desired product N-(β-chloroethyl)-N-methyl-N'-(2-trifluoromethyl-1,3,4-thiaidazol-5-yl)urea as a solid residue

SO MANY reactions!

Extracting our Archive

• What could we get from our archive?• Find chemical names and generate structures• Find chemical images and generate structures• Find reactions• Find data (MP, BP, LogP) and deposit• Find figures and database them• Find spectra (and link to structures)• And of course InChIfy the entire collection

After we mine the Archive

Models published from data

Text-mining Data to compare

Progress to date

• We have text-mined all 21st century articles… >100k articles from 2000-2013

• Marked up with XML and published onto the HTML forms of the articles

• Required multiple iterations based on dictionaries, markup, text mining iterations

• New visualization tools in development – not just chemical names. Add chemical and biomedical terms markup also!

MedChemComm markup

MedChemComm markup

MedChemComm markup

InChIs under our “repository”

• Scientific publications are a summary of work• Is all work reported?• How much science is lost to pruning?• What of value sits in notebooks and is lost?• Publications offering access to “real data”?

• How much data is lost?• How many compounds never reported?• How many syntheses fail or succeed?• How many characterization measurements?

New Repository Architecturedoi: 10.1007/s10822-014-9784-5

What are we building?

• We are building the “RSC Data Repository”

• Containers for compounds, reactions, analytical data, tabular data

• Algorithms for data validation and standardization

• Flexible indexing and search technologies

• A platform for modeling data and hosting existing models and predictive algorithms

New Repository Architecture

Compounds Reactions Spectra Materials Documents

CompoundsAPI

ReactionsAPI

SpectraAPI

MaterialsAPI

DocumentsAPI

CompoundsWidgets

ReactionsWidgets

SpectraWidgets

MaterialsWidgets

DocumentsWidgets

Data tier

Data access tier

User interface

components tier

Analytical Laboratory application

User interface tier

(examples) Electronic Laboratory Notebook

Paid 3rd party integrations (various platforms – SharePoint, Google, etc)

Chemical Inventory application

Deposition of Data

Compounds

Reactions

Analytical data

Crystallography data

InChIs under the repository

• All compound-based data handling will of course connect with InChIs• Compounds• Reactions• Compound-spectra matching • Etc. etc. etc…

For Deposition of Data

• Developing systems that provides feedback to users regarding data quality• Validate/standardize chemical compounds• Check for balanced reactions• Checks spectral data

• EXAMPLE Future work• Properties – compare experimental to pred.• Automated structure verification - NMR

RSC Cheminformatics Projects

• RSC as a provider of support for grant-based projects• Utilizing ChemSpider initially as a platform• Developing Chemical Registry Service• Utilizing core architecture and widgets to

serve the projects

The PharmaSea Website

• ChemSpider IDs and InChIs/InChIKeys made open and available for linking

• Exposed via the Open PHACTS RDF export

• A structure ID standard to enable further linking across the semantic web of science

InChIs and DDP

Electronic Notebook Data

• Development work integrating chemistry into the Southampton Labtrove notebook• Stoichiometry table development• Analytical data integration

• “ChemTrove” includes chemistry widgets and InChI as an important data field

Side Effects of InChI Usage

SMILES by comparison…

Side Effects of InChI Usage

Standardization IssuesDepiction based on molfile

Standardize

• Use the SRS as guidance for standardization• Adjust as necessary to our needs

Nitro groups

Salt and Ionic Bonds

What needs to happen?

• If we could validate• Catch errors in databases (and clean)• Proactively catch errors in publications/patents• Reduce junk in the ether – improve QUALITY!

• If we standardized• Interlinking should improve

Validate and Standardize

CVSP Filtering

CVSP Filtering of DrugBank

DrugBank (ca. 6000 records)

• 38 records with InChI not matching the structure, e.g. DB08521, DB08187

• 24 records where names (IUPAC_NAME) did not match the structure, e.g. DB08346

• 38 records with SMILES not matching the structure, e.g. DB08293

• 53 records with unusual valence, e.g. DB01983 with boron(V)

ChEMBL (1.3 million records)

• 11,020 records with 4 bonds and zero charge, e.g. CHEMBL501101 or CHEMBL501973

• 271 records with hypervalent oxygen (e.g. , CHEMBL2219679), carbon (e.g. 1005895), boron, chlorine, iodine or phosphine

• 6,177 records where direction of bond makes no sense, e.g. CHEMBL12760 and CHEMBL34704

ChemSpider Standardization

• Entire ChemSpider database will be standardized using modified FDA rule set

• Original Molfiles will be standardized and all properties (predicted properties, SMILES, InChIs, Names) will all be regenerated

• CLEAN’ed database to compounds repository

• Standardization procedures automatically applied to all future depositions

Recent Data (last week)

Internet Data

Data Repositories and InChI

Commercial SoftwarePre-competitive Data

Open ScienceOpen DataPublishersEducators

Open DatabasesChemical Vendors

Small organic moleculesUndefined materialsOrganometallicsNanomaterialsPolymersMineralsParticle boundLinks to Biologicals

If InChI was not developed…

• Database linking would suffer dramatically

• The web would not be “structure searchable”

• Cheminformatics tools would likely not be linking to public domain databases in the same way

• We wouldn’t be here discussing….

• And ChemSpider would not have been built

Acknowledgments

• The InChI team

• The entire RSC cheminformatics team…

• Daniel Lowe for the text mining work

• Igor Tetko for OCHEM modeling

Thank you

Email: williamsa@rsc.orgORCID: 0000-0002-2668-4821 Twitter: @ChemConnectorPersonal Blog: www.chemconnector.com SLIDES: www.slideshare.net/AntonyWilliams