The traditional perception of the publishing process has been that it culminates in a print article. The Royal Society of Chemistry (RSC) has for many years been acutely aware that there is a wealth of information contained in scientific communications that we publish and that its true value can only be unlocked by enabling the discovery of the data within them. This is challenging due to the variety of ways that scientists provide data, textually, graphically, and increasingly in supplementary information. This talk will outline how the RSC has applied innovative approaches, developed both internally and externally, to identifying important chemical data within the literature and provides tools to anyone using chemical data to analyse and improve its quality. Examples will include: Project Prospect, the Experimental Data Checker, our CIF data importer, ChemSpider and our structure validation and standardization service.
This presentation was given by David Sharpe at the ACS Fall Meeting 2012 in Philadelphia
2. Overview
• Introduction
– What data can we consider?
– What are the challenges
– What data and sources does the
RSC have?
– Experimental Data Checker
• Case Studies:
– Project Prospect
– Chair forms of
Sugars/cyclohexanes
3. Traditional Chromatography
Images taken from:
http://www.sciencemadness.org/talk
/viewthread.php?tid=3960&page=3
http://en.wikipedia.org/wiki/Column_
chromatography
4. Why Digital Chromatography?
• Useable information is mixed in with description
and analysis – Makes it difficult to find
• Despite our best efforts – still lots of ambiguous
or plain wrong/unusable chemical information
• Why?
– Human error
– Processing errors
– Incorrect usage of data generation/extraction
– Style over meaning
– Data not generated with reuse in mind
– Data generated for humans
5. Style/Layout Vs Meaning
• Structures drawn to illustrate more than
just the identity
• Data not generated with reuse in mind
• Author practices
• Mixed 2D and perspective representations
• Unintentional definition of stereochemistry
6. Data generated for humans
• Separated/Orphaned information inc.
Markush structures, information passed by
reference
7.
8. What chemical data can we consider?
• Chemistry is an especially challenging - wide range of
types of data
– Numeric data
– Names
– Structures
– Terminology
• Over a hugely different set of topics: Org, Inorg,
Physical – Meanings/interpretations are not perfectly
aligned
• Application of standards can be challenging
• Drawing conventions – are documented but not used
12. What is Prospect?
Visible output
Enhanced Prospect InChI–name pairs Better
Output layer RSS
HTML database (in ChemSpider) ontologies
Information layer Enhanced RSC XML
Tool layer OSCAR
InChI–Name pairs Author
Input layer Ontologies RSC XML
(from ChemSpider) CDX files
12
13. People and machines
People Machines
Can understand narratives. Can’t understand narratives.
Can interpret pictures. Can’t interpret pictures.
Can reason about three- Not able to infer 3D structure
dimensional objects. from 2D without cues.
Can do a high-quality job. Can do a lower-quality, but still
useful job.
14. Case study 2: The chair representation
issue
InChI=1S/C6H12O6/c7-1-2-3(8)4(9)5(10)6(11)12-2/h2-11H,1H2
WQZGKKKJIJFFOK-UHFFFAOYSA-N
• 5 stereocentres = 2^5 isomers =32 structures
15. Case study 2: Chair forms of
hexacycles what could go wrong?
16. How do we “fix” chair-representations
How we normalize them:
1. Identify 6-membered rings (Indigo)
2. Identify what sort of ring it is
3. Map atoms onto a standard structure (eg.
beta-D-glucopyranose)
4. Tidy
17. The future: “The digester”
• Ability to:
– Reconnect R-groups
– Expand abbreviations
– Expand brackets
– Link structures with reference IDs
18. Other examples that we didn’t
mention in case studies
• CIF data importer
• Structure Validation and Standardisation
– (Thurs Aug 23, 9:15 am, Marriott Downtown,
Franklin Hall 6)
• Work on creation of ontologies, RXNO, CMO
– Also collaborating on: ChEBI ontology, GO, SO
• Collaboration with Utopia to enable Prospect
mark-up of PDFs
19. Summary
• Many data sharing practices are based on:
– Traditional print articles
– Consumption of data by humans only
• This poses issues for publishers and users alike
• The RSC is developing innovative solutions to
address some of these problems
– Chemical structures are challenging
– Limitations to what a machine methods can achieve
– Need to educate authors to think differently
20. Acknowledgements
• Colin Batchelor - Development and Technical
work
• Jeff White & Aileen Day
• Richard Kidd, Graham McCann and Will
Russell
• RSC ICT staff