Re: Announcement: Bio2RDF 0.3 released

On Mon, Mar 23, 2009 at 12:09 AM, Peter Ansell <ansell.peter@gmail.com> wrote:
> 2009/3/22 Egon Willighagen <egon.willighagen@gmail.com>:
>> On Sun, Mar 22, 2009 at 1:42 AM, Peter Ansell <ansell.peter@gmail.com> wrote:
>>> Do you also provide InChIKey resolution?
>>
>> No. That requires look up, so only works against an existing database.
>> Chemspider is doing this, but is not a general solution. InChIKey's
>> are not unique, though clashes rare, and not observed so far.
>
> I didn't think it required a lookup to derive an InChIKey given an
> InChI.

Ah, sorry. InChIKey can be computed, but I thought you meant resolving
what structure has a given InChIKey... going from InChIKey to
structure does require lookup, generation from InChIKey from structure
(or InChI) does not.

> I realise that clashes are rare but possible, just wondering
> whether it would be supported. Leaving them out altogether just seems
> like missing possibly extra information.

I'll add them where missing.

>>> [1] It is just that InChI's
>>> can get pretty long for complex molecules and it makes it harder for
>>> people to accurately copy and paste them around when needed.
>>
>> Indeed. However, InChIKey is less precise. RDF allowing us to be do
>> things in an exact manner, I rather use InChI.
>>
>>> InChiKey's might be better for general use in RDF because they have a
>>> guaranteed identifier length and therefore won't become cumbersome for
>>> complex molecules.
>>
>> But can never be used for owl:sameAs like relations.
>
> Having them as properties could give someone a quick clue as to
> whether they are looking at the same molecule. Humans do interact with
> RDF (inevitably), and having short hash values can still be valuable.
> Given that hashes are usually designed to amplify small changes, it is
> easier than reading a 10 line InChiKey to determine whether there was
> a difference.

Agreed.

>>> Currently all of the InChI's that I have seen have been as Literals,
>>> but it would be relatively easy to also provide them as URI's to
>>> provide the link since you have a resolver for them set up.
>>
>> That was precisely the reason why I started the service.
>
> Good work.

Thanx for the feedback!

Egon

-- 
Post-doc @ Uppsala University
http://chem-bla-ics.blogspot.com/

Received on Monday, 23 March 2009 06:42:06 UTC