Hi Dave,

On Tue, Nov 22, 2016 at 11:08 AM, David Cosgrove <davidacosgrov...@gmail.com
> wrote:

> I really couldn't decide whether this should just be for you, or to the
> whole list, so I decided to defer the decision to you!

Seems like it's appropriate for the -devel list, so I'm including it on the

> I'm looking at GenerateDepictionMatching3DStructure in
> $RDBASE/rdkit/Chem/AllChem.py.  There seems to be a discrepancy between
> what it says it does and what it actually does.  As far as I can see, the
> reference molecule needs to be the same as the target molecule in terms of
> atom order up to the number of atoms in the target molecule, it generates a
> distance matrix from the given conformation of the reference and that is
> used to make the 2D coordinates for the target.  I can't see the need for a
> depiction for the reference, as the docstring suggests, and the 'piece of
> the molecule is constrained', whilst technically correct (anything in the
> reference after the number of atoms in the target is ignored) isn't doing
> what I thought it would.

Yes, the documentation for that function is pretty badly wrong.

> So, as I move it into the core C++ code, I propose:
> a) Correcting the documentation so that it describes what is happening
> b) Putting some extra checks in, to ensure that, for example, the
> reference molecule has at least as many atoms as the target molecule.
> c) Adding an optional 4th parameter that is a query molecule that will
> allow a mapping of a subset of atoms on the reference onto the target, and
> generating a distance matrix just for those atoms.  That way, you would be
> able to take a set of molecules with a common core that are aligned in 3D
> space and generate a similarly arranged set of 2D depictions.  This would
> not be a breaking change.

These changes all make sense to me.

Rdkit-devel mailing list

Reply via email to