Skip to main content

Interest in Links to Patents From Structures in ChEMBL

By jpo

We are exploring establishing links from the ChEMBL compounds to patents. The implementation can have two basic routes....


  • Links from the interface to patents (simple and quick to do now we have UniChem).
  • Patent uri's in the database itself (more complex, and more difficult to keep up to date, but arguably more useful).


So to help our planning for next year, comments, wishes are most welcome....
Labels:

Comments

Bio to Chem said…
This sounds useful but that would depend on how and what links are going to be made. What would be the source of patent-extracted structures you would match against ?
27 November 2011 at 11:58
jpo said…
Well, ChEMBL is not, and cannot become, a patent database; but there is value in providing links between compounds that are in ChEMBL and the patent literature. The integration would be at the level of proving a link from a ChEMBLid to the underlying patents claiming that compound, simply as a link to the patent document. Initially for compounds, but maybe, depending on how things work out, to targets too.

As to the source of the patent structures. There are a number of initiatives underway at the moment to text-mine chemical structures from patents. We're currently not free to say what some of these sources are, but one source could be the feed from the EPO team.

These structures would be loaded into UniChem (qv) and all the lookups done there.
27 November 2011 at 13:48
Bio to Chem said…
An EPO patent structure feed would link nicely to the EBI patent abstracts and the ChEMBL/UniChem links already in CiteExplore for the papers. The tricky bit is locating the exemplar in the document. The millions of Complex Work Unit-derived structures just surfaced in SCRIPDB might also be might be worth considering but are USPTO-only. For the record you already have indirect patent document links in ChEMBL because the ChemSpider entries have an InChI look-up link to SureChem. You can only open three document links (for free) but some are first-filings. I think I know what one of the other feed options might be but we will see if/when this appears!
4 December 2011 at 14:22
jpo said…
Thanks for the comments. At the moment, we have no funding or resource for any of these, so our aspirations are modest :) Just links to patents from Chemblids.

A big problem with other ways of chemical patent data are shown by your other comments - indirect access through semi-open resources, with significant onus on the user to ensure they don't violate any explicit or ambiguous usage constraints/licenses.

One of the ideas of patent filings is explicitly to make things easy to find so researchers don't waste time recreating other peoples IP, and also can build on top of this. Current systems do not really allow this.....
5 December 2011 at 06:00
Post a Comment