Re: [Cdk-devel] Our ZINC quality assurance data set...

[Egon W. <ego...@gm...>]

On Wed, Aug 13, 2008 at 4:03 PM, Rajarshi Guha <rg...@in...> wrote:
>> If the MINMAX is not possible, we look for alternative means... for
>> example, based on the new substructure fingerprinter...
>
> MinMax is trivial to do in R - just need a set of descriptors. What were
> used originally?

Not sure... is it not mentioned in the CDK News paper?

> Otherwise could do MinMax with the MACCS keys or functional FP fingerprints.
>
>> I'm also wondering if these issues in the ZINC SDF file for the cdk-qa
>> would explain some of the variance in the QSAR descriptor validation
>> results published in CDK News... because each program would recover
>> differently from such problems, reflected in the calculated
>> descriptors...
>
> I'm not sure what you mean

If atom type perception is broken, this affects descriptor
calculation... each different program would perceive atom types
differently if the input is invalid anyway, and therefore, result in
different descriptor values... even if the descriptor algorithm is the
same...

Egon

-- 
----
http://chem-bla-ics.blogspot.com/