Dependencies updates, and better handling of hydrogens, stereochemistry, and aromaticity

[Adafede]

Hi @mmzdouc,

Hopefully this should fix mite-standard/mite_data#533
and the issue you mentioned in the Slack channel

Moreover, it superseeds #65.

I still want to update rdkit and manually fix entries as per #62 but will wait for your input before.

[mmzdouc]

@Adafede thank you for the PR! I have added the test - not passing yet.
Since we are at it, I would propose to remove the network calls to UniProt/NCBI/Wikidata/Rhea to speed up validation - they are throwing warnings only, which are usually ignored anyway. I have already re-implemented them on the frontend (API call to UniProt/NCBI to check the accession before submission, e.g. here) and during release preparation (see here)

[mmzdouc]

(I will push another commit in a minute)

[Adafede]

Passing now 👍🏼

[mmzdouc]

Commit caad15f cuts run time enormoulsy; test are now passing in ca 1 sec (was 35 secs before)

[mmzdouc]

@Adafede I will update some cleanup rules for ketcher and run the new version of mite_extras against our current entries to verify backward compatibility; after that, we are ready to merge I think!

[Adafede]

Let me know if rdkit update should be included or not within this one... if the amount of broken entries will be acceptable or not (anyway, I think the sooner the better or they will just get more)

[mmzdouc]

@Adafede removing R,S stereochemistry for reaction SMARTS seems to work nicely with

echo $smarts_string | tr -d "@"

[Adafede]

@mmzdouc Think I finally got something slightly cleaner...

[mmzdouc]

@Adafede wow all entries are passing 🤩
Out of curiosity, I updated rdkit to the latest version and ran the validation again. 66 entries are failing, mostly due to stereochemistry (expected product vs generated product). For the ones that I have looked at, it seems that the exact enantiomere was produced. I have added a test for one, perhaps you can take another look? I am still in favor of simply flattening the entires and moving on, but if they can be easily salvaged, that would be great, too!

[Adafede]

@mmzdouc

I just made the new version return a warning about stereo mismatches, will help everyone fixing these, tell me if this fits what you expected!

[mmzdouc]

@Adafede great workaround - I like that! I have made some small adujstments to the error text. I would rather not allow people to skip the validation step and I replaced this statement with a shoutout for people to get in touch with the maintainers.
I just saw that a new version of rdkit was released and I ran the validation script. The only entry that does not pass is MITE0000346 which is retired anyway.
For me, this is release-ready. I will do an update of the full stack today.

[mmzdouc]

LGTM

[mmzdouc]

@Adafede we should probably unpin the Python version? We are not bound by rdkit anymore. I am pinning the Python version in the mite_web docker container building anyway

[mmzdouc]

@Adafede I did a quick update of the CHANGELOG. Feel free to replace the Misc changes with a brief description of your most important changes.

[mmzdouc]

@Adafede I fully replaced pre-commit with prek. Will go on and merge the PR. Thanks for all the great work!