On Mon, Apr 15, 2013 at 6:02 PM, JP <[email protected]> wrote: > Hi there RDKitters, > > I was wondering if there is any reason why the feature factory detects > NegIonizable (or PosIonizable) as a feature - but not the actual charges > i.e. Anion (or cation). > > If you are doing feature extraction, to build pharmacophoric models, this > electrostatics data is important. The SMARTS patterns i.e. [+] and [-] and > subsequent fdef definition are trivial, so why aren't these used?
When defining pharmacophoric features, you have to decide whether those definitions will be used for molecules that have been charged (i.e. protonated or deprotonated) already or for molecules that are "uncharged". The first case makes it much easier to define PosIonizable and NegIonizable features (one normally just looks for charged atoms), but requires that you have software that can do the protonation/deprotonation. The second case removes the protonation requirement but requires that you have feature definitions that catch most cases where a site should be protonated/deprotonated. BaseFeatures.fdef follows the "uncharged" approach. It sounds like you want a hybrid behavior. That's pretty easy to get by adding the charge items to the feature definitions. The current definitions only catch positively charged nitrogens. -greg ------------------------------------------------------------------------------ Precog is a next-generation analytics platform capable of advanced analytics on semi-structured data. The platform includes APIs for building apps and a phenomenal toolset for data science. Developers can use our toolset for easy data analysis & visualization. Get a free account! http://www2.precog.com/precogplatform/slashdotnewsletter _______________________________________________ Rdkit-discuss mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
