Improve robustness of restraint search and fix torsional force constant bug

[fjclark]

Is this pull request to fix a bug, or to introduce new functionality?

This PR does two things:

1. Fixes ABFE restraints torsional force constants are too high #359
2. Improves the robustness of the ABFE restraint search to strange user input. I've implemented the improvements discussed here.

I've included the two together to reduce CI runs/ faff, but please let me know if I should create separate PRs in future.

If this is to fix a bug...

This pull request fixes issue #359?

• I confirm that I have merged the latest version of devel into this branch before issuing this pull request (e.g. by running git pull origin devel): [y]
• I confirm that I have permission to release this code under the GPL3 license: [y]

Note that the ordering of candidate anchor points in the restraint search will be slightly different now, as the torsions have less weight when ordering the restraints for selection.

If this introduces new functionality...

Changes proposed in this pull request:

I now:

• Ensure that anchors can never be in the same molecule
• Warn the user if they select the protein/water as the "ligand"
• Warn the user if they supply very few frames for fitting

• I confirm that I have merged the latest version of devel into this branch before issuing this pull request (e.g. by running git pull origin devel): [y]
• I confirm that I have added a test for any new functionality in this pull request: [y]
• I confirm that I have added documentation (e.g. a new tutorial page or detailed guide) for any new functionality in this pull request: [n]
• I confirm that I have permission to release this code under the GPL3 license: [y]

The tests could be more comprehensive - at the moment I've only added a test to check that an error is raised if the receptor and ligand candidate anchor point lists share atoms. I haven't included tests to check that the correct warnings are being raised. The reason for this was that the only trajectory I can find on the website (https://github.com/OpenBioSim/biosimspace_website/tree/main/structures) only includes the ligand and the protein. This means that I can't select weird "ligands" which aren't the ligand or protein (e.g. water) which don't raise errors for other reasons. If we'd like more complete tests, we could always add a trajectory with e.g. 1 water molecule as well as the complex. However, I have verified that these warnings are raised for one of my test systems, and the code for these checks is very simple.

Suggested reviewers:

@lohedges

Thanks.

[lohedges]

Thanks for this. The CI is failing because there isn't yet a 2024.4.0.dev Sire package. I'll trigger the CI for that now.

[lohedges]

Not sure what's going on, but it appears that you might have merged in devel without actually taking the changes, i.e. using ours. This is why the CI is failing, since you are still building against the old version of Sire. (Although the packages should still be available.) For example, this requirements files appears to be the one from main, not devel. I think this might also explain the weird diff with the website index file, which is proposing changes which are already in devel.

[fjclark]

Sorry, not sure what I've done to cause this. Double-checking that my devel was in sync with OBS, pulling again, and merging again changed only requirements.txt:

[lohedges]

No problem. Things look correct now. Ignore the Windows failure, although the CI passed, it crashed when uploading the package to the Anaconda cloud. The Linux 3.12 failure is due to black formatting. (I need to switch to pre-commit at some point so the linting/formatting is automated.)

[lohedges]

Thanks, @fjclark. This looks great and will make the restraint selection a lot more reliable.