Hi Greg,
Thanks for the tips. I had a look at MACCSkeys.py and it doesn't look too bad.
The SMARTS patterns are defined in smartsPatts. What does the 0/1 integer flag
next to the SMARTS string mean? e.g. ('[Si]',0), or ('[#7]',1).
All the best,
George.
> Date: Wed, 4 Mar 2009 19:36:25 +0100
> Subject: Re: [Rdkit-discuss] MACCS Keys C++ Implementation
> From: [email protected]
> To: [email protected]
> CC: [email protected]
>
> Dear George,
>
> On Wed, Mar 4, 2009 at 6:15 PM, George Oakman <[email protected]> wrote:
> > Hi all,
> >
> > I have noticed that the MACCS Fingerprints are only available via the Python
> > framework. Has anyone implemented a C++ version on RDKit already?
>
> Not that I'm aware of, but it would be something useful to have.
>
> >
> > If not, I would be very grateful if someone could give me a few pointers to
> > get me started on implementing a C++ equivalent to
> > the Chem.MACCSKeys.GenMACCSKeys Python function.
>
> Frequently translation of RDKit Python code into C++ is pretty
> straightforward; the datastructures and interfaces usually not
> massively different.
>
> A possible starting point for figuring out how to port the MACCS keys
> is the code for the Crippen descriptors in
> $RDBASE/Code/GraphMol/Descriptors/Crippen.cpp
> That has some of the pieces you would need, including some code for
> reading a set of SMARTS definitions from a text string in a flexible
> manner that one could use as a model.
>
> Once you have the definitions for the MACCS keys, one needs to apply
> them to molecules, count the number of matches, and compare those
> counts to the targets in the parameter set. The fingerprints can be
> returned as an ExplicitBitVect. Note that keys 125 and 166 can't be
> defined purely as SMARTS, but the definitions for these are relatively
> simple.
>
> That's not exactly a recipe for the reimplementation, but it might be
> enough to get you started if you choose to do it.
>
> > This may be silly, but could I simply write a C++ wrapper around the Python
> > function?
>
> you probably could, but I'm guessing that it would be more trouble
> than the reimplementation.
>
> -greg
_________________________________________________________________
View your Twitter and Flickr updates from one place – Learn more!
http://clk.atdmt.com/UKM/go/137984870/direct/01/
