[Dockdev] ZINC structures question

John J. Irwin jji at cgl.ucsf.edu
Mon Sep 15 12:06:05 PDT 2008


Hi Carlos

Thanks for your email.

Carlos P Sosa wrote:
> Hello,
>
> According to the official ZINC paper, the structures defined in ZINC come
> from "suppliers as 2D SDF files".  How accurate are these structures
> compared to experiment?
Catalogs aren't accurate at all, they are just depictions of molecules
in 2D. Thus they aren't "real structures" at all.
>   When adding new groups to this structures, what
> is the recommended way to do it?  via QM calcs?
>   
The structures in ZINC are processed using Corina to make 3D models, and
ligprep, to generate biologically relevant tautomers and protonation
states.  I understand your question to mean, if you wanted to chemically
derivatize structures in ZINC, how should it be done? If that is right,
I would say it depends on the application. But for simply generating
"reasonable 3D structures for modeling" I would consider making the
change in 3D, and then going back to SMILES, and regenerating the
structure using Corina or other similar SMILES to 3D tools.

Of course, you can use QM calculations to refine the geometry of any 3D
structure. However, this would normally be a gas-phase structure, at a
single conformation, so you'll still need to sample its structure in the
binding site, perhaps using QM calcs to refine the docking pose.

If I misunderstood, please feel free to ask again.

John
UCSF ZINC Team


More information about the Dockdev mailing list