Contribution of potentials for coarse-grained PLUM protein model

Hi Andrew,

No problem - It’s a pair style.

I’ll be able to upload my implementation early next week with some documentation.txt. I may just leave a warning regarding the tally ev stuff as Axel said, or try the approach that Steve suggested - not sure yet.

I’m happy to answer all questions.

Gil

I attach files for the PLUM forcefield. The force calculations are well-tested and correct. It would be useful if an experienced developer of LAMMPS checked the ev tallying and implemented the 6-bodied potential’s v tallying. I can provide example input files if that would help.

pair_PLUM_hb.* are the 6-bodied hydrogen bond potentials, which presumably have the complex virial coefficients. I have added a warning on line 100 that the virial calculations are not implemented. Since removing the placeholder virial calculations, I do see divergence from the trajectory on a test system from before I did this, and it would be good to know why.

pair_PLUM_hp.* are simple radial pair potentials, very similar to LJ.

Both are documented in *.txt files.

PLUM_LAMMPS.tar.gz (54.2 KB)

I attach files for the PLUM forcefield. The force calculations are

thanks. i have some comments and questions below from reading through
the doc files.

well-tested and correct. It would be useful if an experienced developer of
LAMMPS checked the ev tallying and implemented the 6-bodied potential's v
tallying. I can provide example input files if that would help.

yes, that would be very much appreciated, especially, if you could
include some comments explaining how to build those inputs or add a
README file with the explanations, so that future users would have it
easier to get started. the deluxe version would be to write a HOWTO
section in the manual, but that might be overkill in this case. a
README is probably best, but already some helpful comments in the
input example would suffice.

pair_PLUM_hb.* are the 6-bodied hydrogen bond potentials, which presumably
have the complex virial coefficients. I have added a warning on line 100
that the virial calculations are not implemented. Since removing the
placeholder virial calculations, I do see divergence from the trajectory on
a test system from before I did this, and it would be good to know why.

i don't understand this. trajectories diverge due to changes to
forces. but any of the code that is used for ev_tally() should not
change forces. can you provide the before/after files?

i would also suggest to adapt the documentation a little bit. i would
suggest to combine the two documentation files and explain that those
potentials only make sense when used with hybrid/overlay pair style
and adapt them accordingly.

as for the identification of which type corresponds to the specific C
atoms and which corresponds to the specific N atoms, i would suggest
to rather specify them in the pair style definition, similar to what
is done for the TIP4P styles identify the water oxygen and hydrogens.

some questions:
- can you explain why "newton on" is required for PLUM/HP?

- i would have expected that there would be mentioning of requiring
particular special_bonds settings or at least an explanation of what
settings are compatible.

- do you mind i we translate the file names and style names to
lowercase to remain consistent with the rest of LAMMPS. by convention,
style names in CAPS signify internal styles that are only to be used
by other styles and thus should not be used directly. for file names,
all lower-case avoids some portability issues with file systems that
are case preserving but case insensitive.

thanks,
      axel.