I created a big chemical database (using flexible docking, pharmacophore modeling methods) of potentially active compounds. Now I have a library involving 10,000 compounds.
I want to recognize similar structures and divide them into groups (there are a lot of simular compounds). Then I’m going to select structures which represent the chemical diversity of each group. I hope it'll be a good final library for biology testing. So, does anybody know any technique or software for such work? I would be very grateful.
If u want realize these features youreself, use Chemistry Development Kit (CDK)
http://sourceforge.net/apps/mediawiki/cdk/index.php?title=Main_Page
Example for searching/highlighting substructure:
http://ctr.wikia.com/wiki/Highlight_a_substructure_in_the_depiction
Have a look at JKlustor
http://www.chemaxon.com/products/jklustor/
I use PaDEL-Descriptor
Available here:
http://padel.nus.edu.sg/software/padeldescriptor/
You will need to get the structures into the SMILE format or something similar.
ChemAxon is good for translating structures.
http://www.chemaxon.com/
PaDEL
I ask for all descriptors to be calculated.
Then run a PCA in R to determine which descriptors actually matter
Run PCA again on the smaller set of descriptors.
You need to develop a clustering task on your dataset :
http://en.wikipedia.org/wiki/Cluster_analysis
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology