Question

Which cheminformatics library is the best to assign 3D coordinates to a structure file, RDKit or CDK? (best quality of the results) I'm looking for a free equivalent for CORINA.

http://www.rdkit.org/ http://sourceforge.net/projects/cdk/ http://openbabel.org/

Answer 1

You can use the Chemical Structure project for the test. It contains ~630 structures. All structures are generated by the ghemical software using a PM3 method. Each structure contains also 1D descriptor string (InChI and SMILES). Therefore, you can compute 3D structure from the 1D descriptor with the different software and compare the results to the Chemical Structure XML file.

Answer 2

I'm not that familiar with CDK's builder3d, and I haven't tried it, but from looking at the code, it seems to use a rule-based or template-based approach for generating an initial guess geometry (for example it seems to use preconstructed ring templates). RDKit, on the other hand, uses a distance geometry approach. I use RDKit and I find it works quite well for my purposes.

[Update] Also, regarding your question about other free tools for 3D structure generation, I am aware of the following:

• OpenBabel (rule-based approach, I think)
• smi23d (distance geometry); used to be at http://www.chembiogrid.org/cheminfo/smi23d/ but it doesn't seem to be available there any longer (anyone (Rajarshi?) know if this is still available?)
• Balloon (distance geometry approach) (discussed previously at http://blueobelisk.stackexchange.com/questions/138/is-balloon-a-free-equivalent-for-corina)
• FROG (rule-based approach) (discussed previously at http://blueobelisk.stackexchange.com/questions/138/is-balloon-a-free-equivalent-for-corina)
• Marvin - http://www.chemaxon.com/marvin/index.html
• PRODRG http://davapc1.bioch.dundee.ac.uk/prodrg/index.html
• BUILD3D (in XdrawChem) (distance geometry approach) http://sourceforge.net/projects/xdrawchem/files/build3d/
• Vconf (free one-year license for academia) http://www.verachem.com/vconf.html

[Update 2]: smi23d can apparently now be accessed at http://cheminfo.wikispaces.com/smi23d

Answer 3

Hello,

http://fiehnlab.ucdavis.edu/staff/kind/ChemoInformatics/Concepts/3D-conformer/

You question, which approach is best to assign 3D coordinates to small molecules: Those programs are the best, that perform best with existing validation sets such as thousands of molecules from X-Ray Crystallography databases.

There is no peer reviewed publication which compares all the latest tools. Just stick with the common accepted ones, aka "When in Rome, do as the Romans do" (BTW that is no platitude, its all about validation)

www.scopus.com/authid/detail.url?authorId=14834041700

Cheers Tobias

Answer 4

Interesting question! We would need an independent test set... perhaps we can set something up based on CrystalEye?