logp help

Hi,





I'm using the logp plugin to get the logPtrue value, and it seems very slow. If I run this against 2000 smiles it takes 77 seconds, compared with another vendor's logp which takes 7 seconds. Am I doing anything wrong ? Tests are run on the same hardware.

















final Molecule mol = MolImporter.importMol(smiles);








plugin = new logPPlugin ();





plugin.setMolecule(mol);





plugin.run();





plugin.getlogPTrue();

Hi,








The logP calculation speed was improved recently. I calculated the logP value of  approx. 50000 molecules using the 5.2 version.   Calculation speed was about 180 molecule/sec. 





May be you have an old version of logP calculator.








Jozsi

The lastest version from the download section on the web is 5.1.05





 http://www.chemaxon.com/download.php?d=/data/download/jchem/jchem-5_1_05-windows.exe





Where can I get 5.2 ?

Hi,








Sorry I misstyped:  The improved speed available from  5.1.4.  So you can download the new version 5.1.05 as you say.  It will be more fast than the older versions. 








Jozsi

Hi,





I'm using the latest jar and I'm not getting any where near 180 Molecules per second more like 22 per sec.





Could you send me the smiles / structure list you use for your test ? (I can't supply you with my list).





Thanks

Hi,





I downloaded smiles file from here :








http://cactus.nci.nih.gov/ncidb2/download.html





(250251 molecules)














Jozsi

With those test smiles I'm getting 180 mol per sec, but our compounds are more complex and I get 29 mol per sec compared to 330 mol per from our other vendor.

Hi,

Quote:

...but our compounds are more complex...

I think many of your compounds contain both acidic and basic groups


as well. This is why anionic, cationic and zwitterionic forms of a compound will be present in the solution at a given pH beside of the unionized form. Distribution of these species depends on the pH, therefore, the experimental value of the apparent logP also should depend on the pH.





One can ask if a given pH will be suitable to get the logP value? How should the distribution of the species be calculated at that pH? And how to calculate the logP in this case?








I wonder how other vendors take the pH dependency of "logP" into account.





Since the logP does not seem to be relevant, for this type of compounds the distribution coefficient logD is calculated at the isoelectric point.





This calculation process is implemented in our logPTrue algorithm which calculates the pKa, isoeelectric point and the logD of a compound.





So, our "molecule/sec" index is not so nice because a large amount of calculation work has to be done to get the correct answers for the above three questions.








Jozsi