charmm force field

[Sizhang92190]

I've been trying to use openfe recently to perform RBFE calculations on a receptor (containing an ADP residue) + a couple of ligands. Due to the presence of ADP, I've had to replace the amber force field with the charmm force field in the script, gufe/settings/models.py.

i.e., 
forcefields: list[str] = [
  "charmm/charmm36_nowaters.xml",
  "charmm/waters_ions_default.xml",
  ] 

However, the calculation failed in the parameterization step due to a memory issue. (It still failed even if the memory was increased.) @mikemhenry Any ideas?
Would appreciate any suggestions! If necessary, I can upload a few files later.

[IAlibay]

Thanks so much for reporting this @Sizhang92190 !

Admittedly, I don't think we've tried using the charmm FFs yet, mostly because of small molecule FF compatibility issues and also I'm not 100% sure that short range shifted cutoffs are working as expected when building the hybrid topology. That being said we should definitely try to support it if we can.

Do you happen to still have the error trace for the memory failure? That would help up try to work out where things are going wrong.

[Sizhang92190]

Thanks for your reply! I guess you're right. There should be FF compatibility issues between charmm FFs for proteins and openff or gaff for small molecules. It then causes the simulation to get stuck in the 'Adding Forces' step when building the hybrid_factory.

[mikemhenry]

Something you can try to debug this issue is to use https://github.com/openmm/openmmforcefields directly to parametrize a system with charmm without having to run things through the openFE pipeline

[IAlibay]

Our approach to cofactors (like ADP) has been to parameterize them via OpenFF as just an additional spectator small molecule. This seems to have worked reasonably well in most of our cases.

[Sizhang92190]

Thanks for the detailed explanation! @IAlibay CMAP atomtypes would also be an issue if I use charmmff with the openfe pipeline.

About using spectator, it's a good idea! But I see that it's now only available in the perses functions that requires an OpenEye license. I don't have one.

Finally, I figured out one way to run RBFE for a system containing ADP and phosphate with the current openfe pipeline. There're two files that need to be modified: 1.) Add funcs that generates bond orders and types of 'nonstandard' residues (e.g., ADP) to
gufe/components/proteincomponent.py, 2.) Use extended_omm_forcefield instead of omm_forcefield in openfe/protocols/openmm_rfe/equil_rfe_methods.py. extended_omm_forcefield contains the generated SMIRNOFF parameters for ADP.

    extended_omm_forcefield=system_generator.forcefield 
    from openff.toolkit import Molecule
    molecules = Molecule.from_file(residue_template_fname)
    from openmmforcefields.generators import SMIRNOFFTemplateGenerator
    smirnoff = SMIRNOFFTemplateGenerator(molecules=molecules)
    smirnoff.add_molecules(molecules)
    extended_omm_forcefield.registerTemplateGenerator(smirnoff.generator)

Since I wasn't sure which openfe or/and gufe functions I needed to modify, I just simply added above code to the second file. (If you have any suggestions, that would be helpful!)

To let the above scripts work, I just need to provide residue_template_fname with a sdf file along with a template residue (e.g., ADP).

Attached are those two modifed files and the template residue file. Just in case if someone has the similar goal. Btw, I use openfe-0.12.0.
files.zip

Please let me know if I make a mistake. Thanks!

[IAlibay]

@Sizhang92190 there's a better way to do this within openfe but it seems like we haven't advertised it very well. It might be a lot easier to discuss in a call if that's something you want to have?

You may need to use a slightly newer version of openfe.

The short story is that you only need to pass through extra SmallMoleculeComponents to your protocol and these will be handled as cofactors.

[Sizhang92190]

An extra SmallMoleculeComponent sounds like a better and simpler way. I really should have known before I tried those things.
I'd like to discuss it with you if you email me an invitation link. (My email: sizhang@temple.edu)

[IAlibay]

Ok, I've just sent you an email with some proposed meeting times.