Re: [Joelib-devel] MolConnZ, import and batch files under windows

[Joerg K. W. <we...@in...>]

Hi Egon,

first: if the e-mail of C. Steinbeck is not correct, please feel free to 
forward this message. Eventually he could be interested in such things 
also. I've read his actual JCICS-CASE paper.

> (Jmol with a lower case m... :)
Jmol :-)

> In other words, what would be most interesting for Jmol to depict?
Import&Export and descriptors, as you surely know i use the OELib 
kernel, also OpenBebl kernel, so we can easily add all supported types 
of them, if anyone can find to port the C++ to Java, which is VERY easy 
... but costs still time ... the important point are the atom types, and 
these are available.

> Anyway, I'm seeing a plugin that would calculate the JOELib descriptors for 
> the shown structure... more ideas? Have a look at 
> http://cdk.sf.net/plugins.html to see what the plugins are about...
I'll have a short look ... and again ...

>>But this does not change my time priorities, so i would be really happy
>>to add any interfaces, but i'm not able to maintain they actively,
>>because my actual focus lies on our interal JCompChem cheminformatics
>>library with data mining and maximum common substructure search
>>algoritms.
> Sure. I have your two recent articles on my desk, but have not found time yet 
> to read them...
Machine learning and algorithm stuff ...

> BTW, CDK already provides MCSS code... why not use that? It has the best 
> algorithm available at this moment... unless you're interesting in finding a 
> better one...
I'm not interested in finding a better one ... i'm interested in using 
them with a general defined approach, so that we can adapt the algorithm 
to our cheminformatics requirements.
The CDK implementation uses, if i'm understanding this correctly the 
association graph method, so i use the same association matrix method 
with a generalized atomType assignment, which contains also the ESTATE 
or my improved CESTATE one ... which works already great using some 
Hashing ...
The association graph is not the problem, but the missing atom types and 
atom properties (descriptors) in CDK.
The relevant clique detection algorithm in my implementation is just an 
interface and can be replaced by any other implementation ... depends on 
your needs, because fast implementations uses, of course, heuristic 
approaches !
BTW, the original reference for this kind of MCS was 1976 !

The performance depends on the used Clique detection algorithm, 
references can be also found in my papers, because i use already two 
implementations, ...there exists a lot of them ... but i do not believe 
that it is possible to improve the performance easily, because there was 
a lot of work already done by graph experts.

We are working on a multiple interpretation and reimplementing three 
other literature approaches, to be able for multiple MCS which is much 
more interesting if you are interested in finding a pharmacophore based 
description based on ligands ... so this is in progress ... or if you 
like some CASE relevant analysing stuff ... at the moment we have 
reached the pre-alpha, so we are testing, testing and testing ... to 
find suitable combinations and parameters ...

Regards, Joerg


> 
> 
>>These things are really alpha, because the packages are
>>refactored yet, because we have found some blind alleys, which restricts
>>further algorithm development, so ... in progress. Eventually these
>>things will be publicly available in the future, too ... as part of
>>JOELib ... and after (hopefully) publishing some nice combinations ...
>>
>>Regards, Joerg
>>
>>P.S.: This reminds me to update the CDK installation instruction ... and
>>add some more instructions to the tutorial ... my actual deprecated CDK
>>version has some problems in 2D layout ...
> 
> 
> Yes, please do keep up... where gaining momentum every month... 
> 
> Egon
> 
> - -- 
> eg...@sc...
> PhD on Molecular Representation in Chemometrics
> Nijmegen University
> http://www.cac.sci.kun.nl/people/egonw/
> GPG: 1024D/D6336BA6
> 
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.0.7 (SunOS)
> 
> iD8DBQFANJ3zd9R8I9Yza6YRAqljAJwLuwh44ehn/Ew6PWNMYU3m7IPXawCaAyDY
> xNHnzDaLo7nWwVYbcDoqi1Y=
> =/HaH
> -----END PGP SIGNATURE-----
> 
> 


-- 
Dipl. Chem. Joerg K. Wegner
Center of Bioinformatics Tuebingen (ZBIT)
Department of Computer Architecture
Univ. Tuebingen, Sand 1, D-72076 Tuebingen, Germany
Phone: (+49/0) 7071 29 78970
Fax: (+49/0) 7071 29 5091
E-Mail: mailto:we...@in...
WWW:    http://www-ra.informatik.uni-tuebingen.de
--
Never mistake motion for action.
                          E. Hemingway