Grouping based on scaffold?

I'm curious,





There are some packages out there that can analyze a compound collection and present a report based on scaffolds, number of members per scaffold, etc. (e.g., SARvision, ChemTK).





Is there anything like this in development for Instant JChem?





Thanks.





- Guillermo.

We certainly want to add powerful analytical capabilites to Instant JChem. This would include:


- R-group deconvolution


- scaffold analysis


- overlap analysis (partially implemented in current version)


- clustering





If you have (or anybody has!) any particular suggestions please could they describe them here in this forum so that that we can include them in our plans.








Many thanks.





Tim

Tim,





I think this will boil down to the approach you decide to implement.





I've used both SARVision and ChemTK and each use a different approach:





1) SARVision: Use a pre-established collection of scaffolds to creat the tree.





2) ChemTK (now sageinformatics.com) uses algorithms to derive the scaffold tree.





SARVision can be faster since it doesn't have to figure out scaffolds. The limitation is when a scaffold is not part of the list then it is not recognized as such.





ChemTK doesn't depend on a pre-established/canned scaffold list, but it could be tricky to select the ideal parameters for the algorithm to catalog/group all compounds members based on common or unique scaffold... and it could require more time and CPU power.





I think SARVision's approach could be used along as the program allows the user to edit/add scaffolds to the list. That way if you happen to use a rare chemotype one you add it to the scaffold list yo're set for good and dont' have to repeat this again.





My 2 cents.





- Guillermo.

My bad, ChemTK now belongs to Simulations Plus.





-Guillermo.

Hi Guillermo,





We already have such tools in the JChem suite, available separately from Instant JChem. There is an maximum common substructure based clustering tool called libMCS. You can check it out here:


Scientific poster: http://www.chemaxon.com/conf/MCS_Based_Hierarchical_Clustering_A4.pdf


User´s guide: http://www.chemaxon.com/jchem/doc/user/LibMCS.html





It is planned to integrate libMCS into Instant JChem in the future, along with R-group decomposition (which is currently only available from command-line and API. See user´s guide: http://www.chemaxon.com/jchem/doc/user/RGroupDecomposition.html )





LibMCS is rather like ChemTK, as it uses the maximum common substructures derived from the library. AFAIK we have no immediate plans to implement the other, scaffold-library approach.





(The discussion for these tools is usually in this part of the forum.)





All the best,


Szabolcs

Hi Szabolcs,





Thanks for the feedback. It's great ot know that LibMCS will be integrated into Instant JChem in the future (near future I hope... :)





I think this approach will work just fine. It's robust, comprehensive, and it doesn't depend on any "canned" list.





This might be a stupid question, but wll all future implementations/additions be avaiblable for Windows, Mac, and Linux?





So far I'm testing the Windows and Linux versions and they rock! ;)





Thanks.





All the best,





- Guillermo.

Yes, we plan to continue to support Windows, Mac and Linux in future versions of Instant JChem.