From: Joerg K. Wegner <wegnerj@in...> - 2005-02-10 10:04:35
especially JOELib2 has much more features than Dragon and we are
transparent at each calculation using the recusrive identifier calculation.
The number of features is a simple calculation: multiply
atomProperty*transformationFeatures if you want (*bondProperty)
transformationFeatures=220;RDF(50*4 smoothing parameters),BCUT(10*2
k-values), you can also apply a PCA before the calculation, so WHIM and
analogue features are not a big deal.
at least: 53*220=11660
So the bottlenck is not the calculation, but the feature selection
problem, but hey there is already the Weka interface.
Unfortunately I have no time to update the tutorial. I have two lectures
at the ACS meeting and going afterwards to the Rauischholzhausen
workshop (Prof. Klebe), so I have no additional time until end of March.
Being more realistic I will have no time until end of June. Sorry!
Any helping hands are welcome ...
Kind regards, Joerg
> On Wednesday 09 February 2005 06:57 pm, Peter Murray-Rust wrote:
>>Are CDK and or JOELib at the stage where they can generate a reasonable
>>number of descriptors?
> Yes, they are.
>>They would have to do a good deal of what, say,
> Combined they can surely match Dragon. The latter has about 1600 descriptors,
> but many of these are basically the same, using just a different parameter
> setting, such at weighting scheme or cutoff radius for radial distribution
> kind of things.
>>And we'd need a fairly rapid starting point - do either have tutorials?
> For CDK, have a look at the command line executable
> Output is CML2, input is MOL files. Subselection should be supported but the
> patch of Rajarshi earlier this week.
> Joerg's tutorial for JOELib is available from:
> which has a "Descriptors" and an "Application" section.
>>If this is of interest we'll continue
> There are others too working on setting up webservices for QSAR... ask on the
> cdk-devel or the qsar ML.
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from real users.
> Discover which products truly live up to the hype. Start reading now.
> Qsar-devel mailing list
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
Never mistake motion for action.
Never mistake action for meaningful action.