Partial charge assignation using NAGL in example notebook

[ijpulidos]

I believe we want to make sure that we use OpenFF's NAGL when assigning partial charges in the python API tutorial. This change accomplishes that.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[IAlibay]

Thanks for doing this @ijpulidos, however it's a little bit more complicated than a straight change.

I believe we want to make sure that we use OpenFF's NAGL when assigning partial charges in the python API tutorial. This change accomplishes that.

This is only partly true for openff-2.3.0 (and above), which is trained with NAGL. It's also a little bit complicated, on our end we don't have a full recommendation on am1bcc vs nagl - indeed it's unclear how it behaves in a lot of system (e.g. net charges), nor do we have a good fallback when it fails (e.g. when you have too small a ligand or atoms that aren't covered by the model).

This is a lot of words to say "let's discuss it at the next protocol devs"?

[ijpulidos]

@IAlibay I see, I agree it's definitely something that's more nuanced, but just wanted to mention here that I needed to change this in order for the tutorial to not error out at all, or else I would get an error (I don't remember exactly which one but I can try reproducing it). Makes me wonder, have you not seen issues running this tutorial in a "modern" (nagl-enabled) environment?

[IAlibay]

@IAlibay I see, I agree it's definitely something that's more nuanced, but just wanted to mention here that I needed to change this in order for the tutorial to not error out at all, or else I would get an error (I don't remember exactly which one but I can try reproducing it). Makes me wonder, have you not seen issues running this tutorial in a "modern" (nagl-enabled) environment?

I don't understand this one - you're saying that ambertools isn't working for you? We've been running it everywhere for a while and haven't had any issues.