parmed can not read Amber prmtop files created by interchange

[steinbt2]

I am trying to process openfftk-created systems in parmed. When writing Amber input files, for some systems the prmtop and rst files created by interchange can not be read by parmed, apparently due to a mismatch of how excluded atoms are being treated.

I have attached a notebook the demonstrates the issue, it seems to happen when ethyl groups are attached to something that is not an sp3-C. Note how toluene runs ok, but ethylbenzene does not.

AmberConversionIssue.zip
condalist.txt

To Reproduce
Run the notebook, observe the traceback

Computing environment (please complete the following information):
RHEL, conda list attached

[mattwthompson]

I'm transferring this issue to the Interchange repo as the bug is not a consequence of toolkit behavior

[mattwthompson]

This looks like #676 which was fixed in release 0.3.2. Could you update your environment and see if the issue persists?

[steinbt2]

I can confirm this, upgrading took me to interchange 0.3.3 and there my example notebook runs through. Do you know if the problem was on the parmed or interchange side? What I mean is, what about prmtop files I wrote with the previous version, which parmed was not able to read. Are those correct or should they be recreated?

[mattwthompson]

Unfortunately, I don't know what the correct structure of a prmtop file is. In general, if I can get Amber to report identical energies to OpenMM and GROMACS, I consider that correct (though that approach is not without pitfalls). I have had a difficult time finding documentation on it, just this PDF from about a decade ago - do you happen to have another reference?

In this case, the behavior difference between ParmEd, OpenMM, sander, and possibly other tools is because some parse the EXCLUDED_ATOMS_LIST and the others re-generate it on the fly.

I encourage you to re-generate all of your files with the most recent version of Interchange and validate the results yourself, I really can't make blanket guarantees about Amber files given the number of options and complications you or any other user could introduce (i.e. any system more interesting than a single ligand in vacuum). If something looks off in your results, please do report back and I will do my best to investigate and patch Interchange in a timely fashion. Please do be aware there are some outstanding issues that may or may not be resolved; I'm working on reaching out to some people in the Amber developers community for help. Some of the current problems are tracked in these threads:

• Unclear relationships between intermolecular exclusions, 1-4 pairs, proper torsions, and improper torsions in Amber files #688
• Incorrect nonbonded energies from Amber if improper torsions are present #686
• vdW energy mismatch between Amber and GROMACS/OpenMM #470
• Drivers: Sander applies SHAKE in single-point energies #323

[steinbt2]

Thanks for the links Matt. I understand that interfacing with Amber with its roots in decades old fortran style inputs can not be easy. I have no better documentation for the prmtop format than the pdf you link and doubt there is any more recent one (apart from chapter 15 of the Amber22 manual). The only people who had to understand it in this detail in the past where probably ones who knew the code inside out, so why document :-/ I assume reaching out to David Case has the highest chance to provide an authoritative answer...