maccs keys

User e34a92cce5

20-09-2006 13:34:33

Hi,


I saw on one of your web pages information on how to write your own virtual screening fingerprint tools using the open API. I am looking to see if I can use the different fingerprint methods to increase my hit rate while running a similarity search. So,for instance, if I have to use MDL Maccs keys(166), how do I go about it? Thanks for your help


Renju

ChemAxon efa1591b5a

20-09-2006 14:37:21

Hi,





the MACCS key implementation is an example code and is not complete, that is, only a portion of keys are implemented. If you intend to use this particular molecular descriptor for your virtual screening experiment you need to complete the implementation.





If, however, you plan to compare other descriptors then you may wish to consider to use some of our tools, like OptimizeMetrics and HitStatistics. http://www.chemaxon.com/jchem/doc/user/ScreeningOptimizer.html





Hope this helps.


Miklos

User e34a92cce5

20-09-2006 16:02:01

Hi Miklos,


Are there resources that you can help me get so that I can get the full implementation. I am not certain whether OptimizeMetrics and HitStatistics will help me since they approach the system more from a virtual screening standpoint rather than just a tool to create fingerprints (Maccs Key).


Renju

ChemAxon efa1591b5a

21-09-2006 16:20:47

Hi Renju,





we could not find a complete definition of the 166 public keys. There is one paper that introduced these keys, though the definitions are vague in some cases. Like those that say "Key X: atom is blahblah or other properties" And these "other properties" are never defined.


Anyway, I can send you the reference tomorrow - I tried to dig out the paper with no success. I have found a more recent one that redefines these keys, and that comes out with 320 keys, and I am not sure that it is more specific than the original publication.


This latter one is: Reoptimization of MDL Keys for Use in Drug Discovery, J.L. Durant at el. JCICS 2002, 42, 1273-1280.





You are right that OptimizeMetrics and HitStatistics will only be useful when the descriptors/fingerprints are readily available for virtual screening, they have nothing to do with the implementation of the Maccs keys (or any other molecular descriptor).





Please let me know if you need the reference to the original Maccs paper, I'll find it for you.





We will be happy to help your efforts to implement the Maccs keys, however, we do not have the resources to do this job for you at the moment. Please feel free to use the example code available in our developers guide at http://www.chemaxon.com/jchem/doc/guide/screen/index.html#custom. This can serve as the basis of your MACCS imlementation. Also note, that the next JChem release (which is coming soon) will provide some explanations too, not just dry code lines.





Regards,


Miklos