Extracting information about which ligand atoms form interactions with protein

[meddwl]

ProLIF has such a wonderful opportunity as generating ligand-protein interactions fingerprint. If this fingerprint can be presented as a dataframe, it is possible to extract information about which residues interact with ligand.
( There is an example how this dataframe looks like https://prolif.readthedocs.io/en/latest/source/modules/interaction-fingerprint.html )
Is there such an opportunity to expand this dataframe and extract information about which ligand atoms or a group of atoms form which type of interaction?

[cbouy]

Hi @meddwl,

Yes you can obtain the indices of atoms responsible for each interaction with df = fp.to_dataframe(return_atoms=True).
The resulting dataframe now contains a pair of values instead of the usual True or False for each interaction.
These pair of values correspond to the indices of atoms in the ligand and protein residues.

For example, if you obtain the following dataframe:

ligand        LIG1.G      ...                                 
protein     ASP129.A      ...
interaction  HBDonor      ...
Frame                     ...                                 
0            (13, 8)      ...

This means that atom 13 in residue "LIG1.G" of your ligand, and atom 8 in residue "ASP129.A" of your protein are making that H-bond interaction in frame #0.

An quick example on how to use this:

atom_idx = 13
resname = "LIG1.G"

lig = u.select_atoms("resname LIG")
lig_mol = plf.Molecule.from_mda(lig)
residue_mol = lig_mol[resname]
atom = residue_mol.GetAtomWithIdx(atom_idx)
coordinates = residue_mol.GetConformer().GetAtomPosition(atom_idx)

Also note that if an interaction wasn't present at a specific frame, you will have (None, None) instead of numbers.

Some examples of scripts to display which atoms are interacting directly on 2D or 3D structures are shown here.