Re: [Rdkit-discuss] Attached: a sdf

[Greg L. <gre...@gm...>]

Hi,

2009/10/26 Christian de Bouillé <amb...@wa...>:
>
>
> Attached 200 chemicals
>> Here a sdf with descriptors computed by JChem from Chemaxon
>> upto <DONOR_COUNT>
>> after the descriptors are calculated by RDkit
>>
>> There are big differences that are the same for different structures

The differences that I see in looking through the first set of
structures, at least in terms of the number of donors and acceptors,
are most likely due to differences in the definitions of the
descriptors. I'm not sure what JChem is using, but if you look in the
file $RDBASE/rdkit/Chem/Lipinski.py you will find the SMARTS
definitions used in the RDKit for acceptors and for donors. If you
disagree with any of these definitions (and it's easy to do so; the
concept is not 100% well defined and certainly is hard to encode
perfectly in SMARTS), you can easily modify the SMARTS to match your
preference.

With respect to PSA (from JChem) or TPSA (from the RDKit), I'm again
not sure what definition they are using. The RDKit definition is based
on the sample code provided by the original authors of the algorithm
and I'm pretty confident that it's correct. Sometimes one sees
differences due to different definitions of aromaticity, but that's
another one of those things that are not well defined.

> I try to use
>>>> from Chem import PyMol
>>>> v = PyMol.MolViewer()
> It does not work
> xmlrpclib is not very clear

This is not well documented, you are right.
You need to have PyMol started and it needs to be running its xmlrpc
server. The command line argument for this is -R

>
> Are you coming to Strasbourg
> 03/11/09
> À 16:00
> Tour de chimie - 1 rue Blaise Pascal - 4e étage - salle informatique 2
> Par le Dr Gregory LANDRUM - NOVARTIS - Bâle
> Is it possible to have a copy of what you will tell ?

Yes, I will be doing a presentation about the RDKit. I will not be
able to give out my slides before the lecture, but I might be able to
get them approved to give out afterwards.

-greg