Dear Greg,

I use these fingerprints:

molsc[i].fp = RDKFingerprintMol(*MolAux1,1,7,2048,2,false);

Are these the same in your benchmark?

Thanks,

Gonzalo

-----Original Message-----
From: Greg Landrum [mailto:greg.land...@gmail.com] 
Sent: 24 July 2012 15:17
To: Gonzalo Colmenarejo-Sanchez
Cc: rdkit-discuss@lists.sourceforge.net
Subject: Re: [Rdkit-discuss] speed of fingerprints calculations

Dear Gonzalo,

On Tue, Jul 24, 2012 at 11:34 AM, Gonzalo Colmenarejo-Sanchez
<gonzalo.2.colmenar...@gsk.com> wrote:
>
> I've been doing speed comparisons between Daylight and the 2012.06 release
> of RDKit for fingerprint calculations. The calculation of a 4179 x 4179
> Tanimoto similarity matrix takes 10 s in DL and 15 s in RDKit. Thus, the new
> RDKit release is much improved in terms of speed, although is still slower
> than DL (by a 1.5 factor).

Even if there's still room for improvement, I'm glad things have gotten better.

> The generation of the fingerprints themselves is about 10 times slower: 4s
> in DL and 42 s in RDKit.

Which fingerprint are you using and with which parameters? The times
seem a bit slow. Again, referring to the benchmarking page
(http://code.google.com/p/rdkit/wiki/Benchmarking): t16 corresponds to
generating RDKit fingerprints for 1000 molecules. This takes about 7
seconds on my MacBook Pro, so I would expect 4K molecules to take
about 28 seconds. This is obviously dependent on the details of the
molecules being considered. An aside: if you want a closer analog to
the daylight fingerprint, you should be sure to set the branchedPaths
argument to RDKFingerprintMol() to false. This excludes branched
subgraphs from the fingerprint and speeds the calculation up a bit.

Best
-greg



------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Rdkit-discuss mailing list
Rdkit-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rdkit-discuss

Reply via email to