Re: [Open Babel] Chirality, etc. [was Re: Babel Rationale]

[Jonathan B. <js...@ca...>]

>As I said elsewhere *holding* this information shouldn't be a problem. The
>questions is what it means, how the concepts are interrelated, etc.
>
>I would suggest that there are the following types of *concept*:
>         - those generally agreed and hopefully formalized (e.g. by IUPAC)

The formalization of rules for chirality descriptors has occurred
orthogonal to IUPAC.  All of the IUPAC recommendations basically boil down
to commentary on the original rules set out by Cahn, Ingold, and Prelog.
So you can say that IUPAC has given the CIP rules their stamp of approval,
but the real formalization happened elsewhere.

>[...]
>
>>o  Two properties for absolute atom and bond CIP configuration,
>>    the first with 7 enumerated values, the second with 4.
>
>It would be very valuable to have these values. Are they all described in
>IUPAC? If not, can you indicate what others there are? (There may be
>excellent reasons for them and it's exactly this that we need for IUPAC to
>amend its dictionaries and rules).

With the exception of a few self-descriptive extensions (like
"Undetermined"), everything is standard.  We provide full references to the
original standard-defining papers.  See
http://sdk.cambridgesoft.com/chemdraw/cdx/properties/Atom_CIPStereochemistry.htm
and
http://sdk.cambridgesoft.com/chemdraw/cdx/properties/Bond_CIPStereochemistry.htm

>We know there is a problem with complete evaluation of CIP values - and I
>know this is an area where you have been personally involved. I am not sure
>at present which software computes CIP and what degree of completeness or
>agreement there is.

ChemDraw's calculations are correct.  Other programs less so, but most
programs that calculate anything are pretty accurate for cases that people
are likely to care about.

>>o  A property for bond wedged- or hashed-ness (at least 14 enumerated values
>>    that are relevent, depending on how you count)
>
>Again this would be extremely valuable. Do authors use these consistently?

Not relevent.  For this property ChemDraw is storing the graphical
representation only, which is guaranteed to be accurate.

>For example I have a recent 2001 paper from a top-quality group in which
>the stereochemical indicator is a *parallel* solid bond. I can only deduce
>the chiral centre because one end is an N. This seems wrong...

I agree.  IUPAC doesn't.  There's at least a third convention.  See my
previous message where I mentioned a new IUPAC project to describe
structure drawing conventions.  The cases where you *can* deduce the
correct chiral center aren't a problem anyway.  You really run into
problems when you get an unwedged bond between two asymmetric centers.

>In another
>part of the diagram there was a zigzag bond - some form of Markush I
>think... So graphical consistency is a problem and guidance from authoring
>tools is important

Agreed.  Each format must describe the ontology it's using, as you have said.

>>o  Two additional properties for hdot/hdash-ness of a fusion atom
>
>What is the currency of these - I thought IUPAC deprecated them?

Don't know; don't care.  People are  still using them.

>If so this
>is the sort of thing that software specs might challenge. (I am not saying
>that everything has to adhere to IUPAC - just that when it does it makes
>life easier)

I prefer to say that IUPAC should adhere to real life...

>>  And we already know of limitations to this model -- the options for
>> describing relative stereocenters are almost certain to be expanded in
>> the future.
>
>Good. Is this driven by CD experience or in response to publications or
>both? Would be very interested in knowing the details.

Both, plus user feedback.  We like user feedback.  We'll have details once
we've figured them out ourselves.


Jonathan Brecher
CambridgeSoft Corporation
js...@ca...