Opposite partial charge interaction

[lanselibai]

I am interested in identifying the opposite charge interactions between two atoms (e.g. within 0.45 nm) from ligand and protein. It looks like Anionic and Cationic interactions only consider fully charged residues like -NH3+, COO-.

The atom.charge contains the partial charge of an atom. How to include this in the plf?

---

I think I got it work by using the code below

# === 1) Define your custom OppCharge class ===
# (Assuming ProLIF ≥ 2.x; change import if you’re on v1.x)
from prolif.interactions.base import Distance

class OppCharge(Distance):
    """
    Custom distance-based interaction that flags any ligand–protein atom pair
    whose partial charges have opposite sign and are within 4.5 Å.
    """
    def __init__(self, distance: float = 4.5):
        # Use "*" SMARTS to match all atoms; will filter by charge sign later
        super().__init__(lig_pattern="*", prot_pattern="*", distance=distance)
        self.name = "OppCharge"

    def atom_pair_mask(self, ligand, receptor):
        base_mask = super().atom_pair_mask(ligand, receptor)
        # Build NumPy arrays of partial charges
        q_lig = np.array([atom.charge for atom in ligand.atoms])
        q_rec = np.array([atom.charge for atom in receptor.atoms])
        # True where product < 0 (opposite signs)
        opp_charge = (q_lig[:, None] * q_rec[None, :]) < 0
        return base_mask & opp_charge

    def detect(self, ligand_residue, protein_residue):
        return super().detect(ligand_residue, protein_residue)

# === 2) Inject OppCharge into prolif.interactions ===
import prolif.interactions
prolif.interactions.OppCharge = OppCharge

# === 3) Extract default interaction names ===
template_fp = Fingerprint()
default_names = list(template_fp.interactions.keys())
# default_names == ["Hydrophobic", "HBDonor", "HBAcceptor", "PiStacking",
#                   "Anionic", "Cationic", "CationPi", "PiCation", "VdWContact"]

# === 4) Append "OppCharge" to that list ===
default_names.append("OppCharge")

# === 5) Build a new Fingerprint with the extended list of names ===
fp = Fingerprint(interactions=default_names)

# === 6) Verify that OppCharge appears in the new fingerprint’s internal dict ===
print("Available interactions in fp.interactions:")
for name in fp.interactions:
    print("  ", name)
# Should list: Hydrophobic, HBDonor, HBAcceptor, PiStacking,
#              Anionic, Cationic, CationPi, PiCation, VdWContact, OppCharge


all_names = plf.Fingerprint.list_available()
# all_names might be:
# ["Anionic", "CationPi", "Cationic", "EdgeToFace", "FaceToFace",
#  "HBAcceptor", "HBDonor", "Hydrophobic", "MetalAcceptor",
#  "MetalDonor", "OppCharge", "PiCation", "PiStacking", "VdWContact",
#  "XBAcceptor", "XBDonor"]

fp_all = plf.Fingerprint(interactions=all_names)
fp = fp_all

[cbouy]

Hi,

Was this code-snippet AI-generated? Not sure where that atom_pair_mask comes from.
Also you won't need this line, it doesn't do anything:

prolif.interactions.OppCharge = OppCharge

ProLIF knows you're adding a new distance-based interaction as soon as you use class XXX(Distance):

Anyway, ProLIF uses RDKit molecules under-the-hood in the interactions, so the charges attribute doesn't exist like in MDAnalysis, but it is available through atom.GetDoubleProp("_MDAnalysis_charge") for each atom in the rdkit mol.

Here's some code based on that (untested)

from math import prod

class OppCharge(Distance):
    def __init__(self, distance=4.5):
        super().__init__(lig_pattern="*", prot_pattern="*", distance=distance)

    @staticmethod
    def get_charges(metadata, lig_res, prot_res):
        return [
            res.GetAtomWithIdx(metadata["indices"][component][0]).GetDoubleProp(
                "_MDAnalysis_charge"
            )
            for (res, component) in [(lig_res, "ligand"), (prot_res, "protein")]
        ]

    def detect(self, lig_res, prot_res):
        for metadata in super().detect(lig_res, prot_res):
            charges = self.get_charges(metadata, lig_res, prot_res)
            if prod(charges) < 0:
                metadata["charges"] = {"ligand": charges[0], "protein": charges[1]}
                yield metadata

[lanselibai]

Thank you very much!

Yes, my code was AI-generated and after some minor debugs it gave me sensible results.

Your code yields very similar results as mine, only very slightly lower values, but negligible.

I will use your "official" one 👍