Re: [Cdk-devel] put in order qsarTests

[Egon W. <ewi...@un...>]

On Monday 30 October 2006 08:39, ten...@gm... wrote:
> To make this point clear:
> XlogP is not based on a predictive model, it sums up contributions of
> atomtypes similar to the Ghose and Crippen logP. The only point here would
> be to discuss if its atom perception should be extracted and be put into
> the atomtypes package.

Christian,

yes, please do! I have started doing this for other atom type perception 
algorithms, and started setting up an elaborate JUnit testing (well, not so 
elaborate yet, but...)

There are two interfaces: one that makes an exact match, and expects 
everything to be set properly (bond orders, hydrogen counts, ether implicit 
or explicit, charges, etc), and one that is more flexible which returns a 
list of possibly matching atom types, leaving room for unset properties, like 
missing bond orders...

I think it is important to have the atom type perception separately so that we 
can thoroughly test it, and run it on itself, which is at least important for 
educational purposes.

Egon



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