Dear mark,
many thanks for taking the time to look at the issue.
It will be interesting if someone speedup this part of the code. You can image
if we call this descriptor (partial charge calculation) for 1 million of
molecules like we do all the time, this will take at least 2 years on your
computer (considering 1.5 minutes per molecule).
I have used some commercial packages like PETRA or MOE which take a fraction of
a second for the molecule in question. PETRA and MOE as well as CDK are using
the same sigma-pi partial charge calculation algorithm, am I wrong ?.
thanks anyway for your help.
Aziz
***************************************************************************
Nova Decision
Products & Services in Molecular Informatics & Chemo-computing
for drug discovery
29 rue de Navacelles 34090
MONTPELLIER Cedex - France
Tel 1 : +33 4 67 41 77 10
Tel 2 : +33 6 60 60 92 01
Fax : +33 4 67 41 79 13
http://www.novadecision.com
****************************************************************************
> Date: Tue, 24 Nov 2009 14:45:22 +0000
> From: ma...@ebi.ac.uk
> To: aya...@hotmail.fr
> CC: cdk-user@lists.sourceforge.net
> Subject: Re: [Cdk-user] Cdk-user Digest, Vol 42, Issue 13
>
>
> Hi Aziz,
>
> What I could see that on my computer with the given SMILES that it takes
> about 1.5 minutes to run in Eclipse, so a lot shorter than 15 minutes.
> The main bottleneck seems to be this bit:
>
> /*processing for which bonds which are not in resonance*/
> for(int number = 0; number < ac.getBondCount() ; number++){
> IAtomContainer aa = setAntiFlags(container,ac, number,true);
> if(aa != null){
> IMoleculeSet ab = gR2.getStructures((IMolecule) aa);
>
> For each bond, setAntiFlags is called and StructureResonanceGenerator's
> method getStructures is called. Each call can take seconds, and that
> adds up to more than a minute and a half.
>
> I can't be of much more help, perhaps Miguel Rojas would have some
> pointers how to possibly speed things up.
>
> cheers,
> Mark
>
>
>
> > Dear Mark,
> >
> > > Could you send the source code that runs for 15 minutes to the list? It
> > > should be relatively easy to debug it, to find out where things stall.
> >
> > Thank you for lookng at this issue. here is the jython code I am using
> > to calculate partial charge for an atom.
> >
> >
> > sp = smilesParser()
> >
> > smi =
> > /"COC1=CC=C(c...@h]2n(C)C(=O)[...@h](C)NC(=O)[C@@H](C)NC(=O)[...@h]3cc4=cc=c(O)C(OC5=CC=C(C=C5)C(O)[...@h](N(C)C(=O)[...@h](C)NC2=O)C(=O)N3C)=C4)C=C1"/
> >
> > mol = sp.parseSimles(smi)
> >
> > mol = reader.next()
> >
> > desc = PartialTChargePEOEDescriptor ()
> >
> > atoms = mol.atoms().iterator()
> >
> > firstAtom = atoms.next()
> >
> > pcharge = desc.calculate(firstAtom, mol).getValue().doubleValue()
> >
> >
> >
> >
> >
> > I also went through the code of of the one of the java object that is
> > instanced by PartialTChargePEOEDescriptor and one of its methods that
> > is called to calculate the partial charges, I found out some bizare code
> > I paste here:
> >
> >
> >
> > public class GasteigerMarsiliPartialCharges implements IChargeCalculator {
> >
> > …
> >
> > public IAtomContainer
> > assignGasteigerMarsiliSigmaPartialCharges(IAtomContainer ac, boolean
> > setCharge) throws Exception {
> >
> > ….
> >
> > double[] q_old = new double[ac.getAtomCount()];
> >
> > for(int i = 0 ; i < q_old.length ; i++)
> >
> > q_old[*0*] = 20.0; ////// STRANGE
> >
> > ….
> >
> > the index should "i" not "0" if I understand this part of the code.
> >
> >
> > thanks ,
> >
> >
> > AYASRI
> >
>
_________________________________________________________________
Tchattez en direct en en vidéo avec vos amis !
http://www.windowslive.fr/messenger/------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now. http://p.sf.net/sfu/bobj-july
_______________________________________________
Cdk-user mailing list
Cdk-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/cdk-user
