ChemSpider users will likely have seen InChIs exposed on the record view for any chemical structure. The majority of you will know what an InChI is..for those of you who don’t I recommend seeing the Wikipedia description. This week the support of InChI was extended by introducing the InChI key in the version 1.02 beta release.

The definition of the InChI key is given as:

“A fixed-length (25-character) condensed digital representation of the Identifier to be known as InChIKey. In particular, this will

* facilitate web searching, previously complicated by unpredictable breaking of InChI character strings by search engines
* allow development of a web-based InChI lookup service
* permit an InChI representation to be stored in fixed length fields
* make chemical structure database indexing easier
* allow verification of InChI strings after network transmission.

An example of InChI with its InChKey equivalent is shown below. There is a finite, but very small probability of finding two structures with the same InChIKey. For duplication of only the first block of 14 characters this is 1.3% in 109, equivalent to a single collision in one of 75 databases of 109 compounds each.

Caffeine:
InChI=1/C8H10N4O2/c1-10-4-9-6-5(10)7(13)12(3)8(14)11(6)2/h4H,1-3H3
InChIKey=RYYVLZVUVIJVGH-UHFFFAOYAW
First block (14 letters), encodes molecular skeleton (connectivity): RYYVLZVUVIJVGH
Second block (8 letters), encodes proton positions (tautomers), stereochemistry, isotopes, reconnected layer: UHFFFAOY
Flag character, indicates InChI version, presence/absence of fixed H layer, isotopes, and stereochemistry: A
Check character: W”

Egon Willinghagen, as prolific as ever (!) has already provided feedback on the beta version on his blog. We have been waiting for the InChI key for a while since we heard of of its development. We’ve been given some grief about the length of some of the InChIs for the molecules on ChemSpider.For example, see this record – the InChI is:

InChI=1/C74H114O37/c1-17-27-94-36-19-23-38(24-20-36)96-34-45-52-59(85-8)67
(93-16)74(104-45)109-51-44(33-79)100-70(63(89-12)58(51)84-7)107-49-42(31-77)
102-72(65(91-14)56(49)82-5)111-53-46(35-97-39-25-21-37(22-26-39)95-28-18-2)
103-73(66(92-15)60(53)86-9)108-50-43(32-78)99-69(62(88-11)57(50)83-6)105-47-40
(29-75)98-68(61(87-10)54(47)80-3)106-48-41(30-76)101-71(110-52)64(90-13)55(48)
81-4/h17-26,40-79H,1-2,27-35H2,3-16H3/t40-,41+,42-,43-,44+,45+,46-,47+,48+,49+,
50+,51+,52+,53+,54-,55+,56-,57-,58-,59-,60-,61+,62+,63+,64-,65+,66+,67-,68+,69+,
70+,71+,72+,73+,74+/m0/s1
(This InChI is word-wrapped to allow it to stay within the WordPress display)

and the InChI key is: CINKENNKQKVJAW-LJIFKLEWBD

Which do you prefer?

Some of you might have read about my analysis about erectile dyscfunction and how the size of the InChI does give a bigger SMILES in general. This IS an R-rated blog-posting so maybe not… Now, with the InChI key all of these issues will be resolved for different erectile dysfunction drugs…everything will now be the same length.

The ChemSpider database online has expanded to over 17.2 Million compounds. The last record in the database is 17628589….there are no properties yet generated for the last 800,000 structures added and the zoom view is not yet available….these are coming shortly.

We have already downloaded the InChi 1.02 beta version and in a few hours have generated and populated the database with 17.2 million InChi Keys. This coming week we will check for duplicates as part of our validation process. What we can say so far is that the generation process was fast and efficient. More reports to come…now, if anyone out there from Google wants to point their indexing of InChI keys our way feel free!

Stumble it!

12 Responses to “The InChI Key Has Been Unveiled – and now there are 17.2 MILLION of them online”

  1. Egon Willighagen says:

    “More reports to come…”

    Yes, please report if you find key clashes. Also interesting is to look at the key densities in key space? Are the generated keys randomly scattered over all possible keys, or is there some kind of preferred key string.

  2. Joerg Kurt Wegner says:

    I would love to see a statistical analysis about the InChI keys. And if possible, please forward the clashes, if any, to the IUPAC. They can try to work on this ;-)

    In fact, I remember a very nice paper where they did something similar. But this was a loooong time ago.

    Cheers, Joerg

  3. Egon Willighagen says:

    On the CCL list it is reported that the major open compound databases, like ZINC, PubChem, and no clashes have been found.

  4. Antony Williams says:

    We have not yet checked for clashes but based on other experiences I doubt we will find them..and if we do it may be more indicative of lapses in our de-duplication procedures and therefore be very useful. We WILL do this work shortly and report back if we find any issues.

    Joerg – can you comment more on the “statistical analysis”?

  5. Joerg Kurt Wegner says:

    A brief answer about the hashing problem was given on the CCL list today (eh, yesterday)
    http://www.ccl.net/cgi-bin/ccl/message-new?2007+09+10+001

    The paper I had in mind was written already in 1994 by WDI, the multi-lingual cheminformatics guru ;-)
    Article (ig94) Ihlenfeldt, W. D. & Gasteiger, J. Hash Codes for the Identification and Classification of Molecular Structure Elements J. Comp. Chem., 1994, 15, 793-813.

    In general, any hashing identifier might run into collision problems, this is exactly a reason why they have to be designed. The purpose of hashing is a fast indexing and reducing the collitions, not necessarily to avoid them completely. In other words, creating globally optimal hashing identifiers is an optimization or statistics problem, which has to be solved by some smart computer scientists or mathematicians.

  6. ChemSpider News » Blog Archive » InCHI keys Now Available on ChemSpider says:

    [...] InChI Key has recently been released in a beta form (version 1.02) and is discussed in more detail over on the ChemSpider Blog. As of today these InChIs have been made searchable on the ChemSpider [...]

  7. Antony Williams says:

    We have now checked for clashes. We do not see ANY clashes at all – there are no two hash keys representing different structures. We DID find some non-deduplicated structures and these will be resolved shortly.

    InChIKey is working as expected.

  8. Antony Williams says:

    The InChiKey has now been linked up to perform direct searched on Google… see http://www.chemspider.com/news/?p=76

  9. ChemSpider Blog » Blog Archive » Seth Godin’s Big Ideas, the InChiKey and Structure Searching the Web says:

    [...] A few years ago at a meeting in Washington I sat in on probably the earliest public forum discussion on the potential of InChI. As a result of excellent teamwork between NIST and IUPAC, and doing rather than just talking they got it done. There was some negativity expressed during the initial meetings about InChI but it did not distract the team from producing the prototype versions, initial release and now the latest update with InChIKey support. [...]

  10. ChemSpider Blog » Blog Archive » Access to ChemSpider Web Services Starts - Initial Exposure of InChI Related Services says:

    [...] about that soon). As part of this general direction and because of the recent interest in InChI (1,2) we decided to expose a series of web services built on a combination of OpenBabel and the InChI [...]

  11. Henning Hermjakob says:

    Is it possible to convert from an InChI key to a molfile?
    I know there’s a method on http://www.chemspider.com/inchi.asmx which does that based on lookup of InChI keys in Chemspider. But I understand there is no direct algorithm to do this, as the InChI key generation is a one way transformation?

  12. Antony Williams says:

    the only way to convert from an InChIKey to a molfile is via a lookup. The InChIString can be converted though.

    it IS a one way generation

Leave a Reply