Incorrect Saltbridge identification on Disaccharides

[ct-jcabot]

Describe the bug
plip.structure.preparation.py identifies as carboxylates substructures that it shouldn't

To Reproduce
By running plip -i 2rfy -yv we can see that plip caractherizes as negatively charged substructures that it shouldn't (R-O-C-O-R and R-O-C-O-H groups)

Expected behavior
These groups shouldn't be negatively charged and therefore no Saltbridges should be identified in this case

Screenshots

As for my tests the problem arrises in lines 770-772 of preparation.py, where the simple Carbon bound to two Oxygens check is not robust enough, i have fixed the issue locally with this code as replacement

if group == 'carboxylate':

  carboxyl_smarts = pybel.Smarts("[CX3](=[OX1])[OX1-,OX2H1]")

  mol = pybel.Molecule(atom.OBAtom.GetParent())

   if any(match[0] == atom.idx for match in carboxyl_smarts.findall(mol)):
        return True
   else:
        return False

I've seen that instead of SMARTS this code uses other rules to define functional groups so a different approach may be better suited for this codebase

[Aroids]

Thank you for including your patch, ct-jcabot.

To the PLIP collaborators:

I too am having the issue of PLIP identifying salt bridges involving ligand carboxylate groups despite no carboxylate being present. In my case, an ester is being classified as a carboxylate.

The chunk defining carboxylates in preparation.py, which ct-jcabot references above, seems to only require a central carbon bound to two oxygen atoms and one carbon atom— is that correct?

if group in ['carboxylate', 'guanidine'] and atom.atomicnum == 6:  
   if n_atoms.count(8) == 2 and n_atoms.count(6) == 1:

I feel like I might be overlooking something, because I don't see how the above alone could distinguish between carboxylic acids, carboxylates, esters, acyl phosphates, etc.

[Aroids]

@ct-jcabot, would you mind clarifying why you suggest using

[CX3](=[OX1])[OX1-,OX2H1]

as opposed to excluding carboxylic acids using

[CX3](=[OX1])[OX1-]

[ct-jcabot]

@Aroids The main thought behind using [CX3](=[OX1])[OX1-,OX2H1] instead of the stricter [CX3](=[OX1])[OX1-] was to ensure that carboxyl groups are identified regardless of the ligand's initial protonation state in the input file.

Since most ligands with carboxylic acids I work with exist primarily in the deprotonated form at physiological pH, this SMARTS allows PLIP to recognize the functional group even if the input structure is protonated. There is certainly room to discuss whether PLIP should strictly respect the input's protonation state or if it should allow for 'dynamic' identification of these groups to account for pH-dependent states.