Pat Riley
Pat Riley
[issue_8458.zip](https://github.com/user-attachments/files/19826506/issue_8458.zip)
Keeping this open because I am still hoping to revisit this and push a fix.
Is there any developer familiar with this that can point me in the right direction?
I filed an issue on their repo, but haven't heard anything. https://github.com/ncbi/pubchem-align3d/issues/2 Do you have other contact info?
Amazing debugging @DavidACosgrove. My understanding of the intention of the algorithm is that no matter the initial pose of the molecules, the starting positions for optimization should be exactly the...
@bp-kelley : for symmetric (or even almost symmetric due to numeric noise), you can get flips along any of the axes. That's what you mean, correct?
> I thought I'd document here, as I haven't seen it anywhere else, that it seems to be optimising only the shape fit, and scoring the features/colours at the end....
> I think this need to be documented and the default changed to 0.5 Just made this PR https://github.com/rdkit/rdkit/pull/8724 but as I say there, I haven't gotten my dev environment...
@DavidACosgrove : thanks for the pointer about doing canonicalization. I can confirm that adding this canonicalization (for reference and probe) before calling rdShapeAlign makes it *much* more reproducible in my...
I'm sure you are thinking about this, but I had to do a little bit a bookkeeping to still align back to the original configuration ``` # Canonicalize here. The...