Addressing issues 75 & 81

[bobym]

1. Updating MCL1 protein.pdb to have caps.

2. Replacing all ligands.sdf with new versions to fix the ligand distortions seen in issue Abnormal ligand conformations #81

3. Adding re-curated thrombin ligands (expanded ligand set)

[IAlibay]

I visually inspected all the ligands, can confirm there are no further abnormal conformations.

[IAlibay]

@bobym, the #75 fixes don't seem to be here?

[bobym]

@IAlibay sorry this got overwritten when I merged my two PRs. Should be there now!

[ijpulidos]

I'm not sure this needs to be addressed, but I see some .reflig.maegz files that are not plain text files, and if these keep changing overtime they could end up consuming a lot of space in the repo. I wonder if we want these at all and if so, maybe we would need to add them as artifacts. What would be the best approach for this?

[IAlibay]

Wasn't too sure on the convention for smiles, so I just used a plain rdkit:

sdf_supp = Chem.SDMolSupplier('02_ligands/ligands.sdf', removeHs=False)
sdfs = [m for m in sdf_supp]

for sdf in sdfs:
    smiles = Chem.MolToSmiles(sdf)
    name = sdf.GetProp('_Name')
    print(f'{name}: {smiles}')

type of approach. I'm not sure if we want the smiles to be isomeric (we seem to have different isomers for some entries I think?) or explicit bonds/Hs, etc... the whole thing seems rather undefined.

[IAlibay]

I'm not sure this needs to be addressed, but I see some .reflig.maegz files that are not plain text files, and if these keep changing overtime they could end up consuming a lot of space in the repo. I wonder if we want these at all and if so, maybe we would need to add them as artifacts. What would be the best approach for this?

As per today's call - maegz files need to be removed from the git history and added in as a artifacts.

[ijpulidos]

While trying to run the structures using perses I faced the following issue for the pfkfb3 target

ValueError: No template found for residue 449 (POP).  This might mean your input topology is missing some atoms or bonds, or possibly that you are using the wrong force field.

Which refers to the POP residue in

protein-ligand-benchmark/data/pfkfb3/01_protein/crd/protein.pdb

Lines 7765 to 7773 in d6ad2bf

	HETATM 7261 P1 POP A 602 92.394 47.635 260.369 1.00 25.65 P
	HETATM 7262 O1 POP A 602 91.918 47.441 258.922 1.00 24.48 O
	HETATM 7263 O2 POP A 602 91.938 48.980 260.962 1.00 23.09 O1-
	HETATM 7264 O3 POP A 602 92.086 46.427 261.263 1.00 22.53 O1-
	HETATM 7265 O POP A 602 94.120 47.663 260.320 1.00 21.12 O
	HETATM 7266 P2 POP A 602 95.359 48.505 259.472 1.00 22.73 P
	HETATM 7267 O4 POP A 602 95.942 47.517 258.447 1.00 23.19 O
	HETATM 7268 O5 POP A 602 96.375 48.888 260.559 1.00 19.27 O1-
	HETATM 7269 O6 POP A 602 94.699 49.715 258.803 1.00 22.72 O1-

[IAlibay]

@ijpulidos I got confused between PRs 🤣 - yeah we should just fix this here whilst we're at it.

[IAlibay]

@ijpulidos I got confused between PRs 🤣 - yeah we should just fix this here whilst we're at it.

I manually extracted the POP from the PDB and placed it in the cofactors.sdf in 2784fc7

[codecov]

Codecov Report

Merging #82 (c416bf2) into main (7da6cae) will decrease coverage by 0.08%.
The diff coverage is n/a.

Additional details and impacted files

[IAlibay]

As far as I can tell, the way this used to work is that it somehow seemed to assume that rdkit molecules hashed to the same thing. Since this is all likely going away with #78 I'm going to say we can just temporarily disable it for now (it's not related to this PR)?

[ijpulidos]

Apologies for the delay on this review. This looks good to me! The structures are getting run as far as I can test with perses. We need to re-check with the changes in #83 once edges are regenerated.

[IAlibay]

Just trying to fix CI and then I'll squash merge.

[IAlibay]

Excessive RTD memory consumption is a new one to me :/ Will deal with this elsewhere, probably need to switch to mamba.

[IAlibay]

Squash merging to avoid large commit diffs.