Support for hybrid molecules (including dummy atoms)

[noahharrison64]

Hi ProLIF team,

Thank you for the amazing tool — it’s been incredibly helpful across a range of projects!

I have a question regarding the compatibility of hybrid molecules with ProLIF.
In the context of non-equilibrium switching free energy calculations we simulate hybrid ligands (e.g., generated via PMX) in GROMACS. Each ligand pair undergoes an end state equilibration prior to short transitions between lambda states. I'd like to try and plot the protein-ligand interactions during these end state equilibrations. The hybrid topologies in these states include dummy ligand atoms which are non-interacting with the rest of system (other than bonded interactions to the rest of the ligand molecule).

My typical setup:

• Hybrid .gro and .top files generated via PMX
• Trajectories simulated via GROMACS creating xtc and tpr files
• MDAnalysis is used to read these trajectories
• Prolif attempts to convert the residues/molecules into rdkit mols

This obviously fails due to the presence of dummy atoms and bonds leading to invalid/non-sanitizable molecules. I'm wondering if you have ever figured out whether it'd be possible to modify the standard Prolif process to deal with dummy atoms in the context of free energy calculations?

Thanks in advance for any guidance or thoughts!

Best,
Noah

[cbouy]

Hi Noah,

If I manage to get this PR merged in MDAnalysis for their 2.10.0 release, you'll be able to bypass the MDAnalysis bit of code that tries to infer bond orders for your dummy molecule (and fail) and just provide your own code to apply bond orders/charges to the molecule produced from the topology.
There are some snippets here that you could use in the mean time for inspiration but it's not great as you can't use multiprocessing on top for some reason