object has no attribute 'gcmc_status' / NaN energies

[Maxim-93]

Hi there!

Running into a couple of issues here when trying to run your equilibration section for the BPTI simulations,

When executing file equil-uvt1.py I get the error message:
Particle coordinate is nan.
I tried changing the integration timestep to 1fs but still see the same result.

Additionally, after just integrating over a few hundred steps (which can be done successfully) I tried to run this block of code:

gcmc_ids = np.where(gcmc_mover.gcmc_status == 0)[0]

but get the error message:
'StandardGCMCSphereSampler' object has no attribute 'gcmc_status'.

Any idea how I might obtain and write out the ghost particles here? Do I just need their respective indices as produced by gcmc_mover.ghost_file (bpti-uvt1-ghosts.txt is the output in this case).

Or should I use
gcmc_mover.deleteGhostWaters
and then save the file resulting from:
positions = simulation.context.getState(getPositions=True, enforcePeriodicBox=True).getPositions()
as a pdb?

[Maxim-93]

For some added context, I see the following after doing:

gcmc_mover.move(simulation.context, 10000)

for i in range(100):
    gcmc_mover.move(simulation.context, 1000)
    gcmc_mover.report(simulation)
    simulation.step(5)

21000 move(s) completed (4 accepted (0.0190 %)). Current N = 0. Average N = 0.982
INFO:grand.samplers:21000 move(s) completed (4 accepted (0.0190 %)). Current N = 0. Average N = 0.982
INFO:grand.samplers:Change in N of -2 between GCMC batches

[olliemelling]

Hi,

The examples in the grand-paper repo are for reproducing the results in this manuscript: https://pubs.acs.org/doi/full/10.1021/acs.jcim.0c00648 and therefore they only work with the versions of grand and OpenMM that were used at the time the data was generated. If you want examples of using the latest version of grand then have a look here.

If you do want to run the specific examples in this repo then I would suggest creating a new conda environment and installing grand 1.0.1 and OpenMM 7.4.2 as per the commands below. The scripts should then run fine.

conda install -c omnia openmm=7.4.2
conda install -c anaconda -c essexlab -c schrodinger -c conda-forge grand=1.0.1

This should fix the exploding you were seeing and also allow you to access the waters in the same way you were trying to (how we tracked the status of each water was changed in the most recent version).

We use "ghost" waters to refer to any water that is currently completely decoupled from the system and therefore not interacting with anything else - is it definitely these waters whose positions you want to write out?

It is common to remove the ghost waters and write out the positions of all non-ghost particles - this can be done using the code at the end of the equil-uvt1.py script, which should now work for you if you use v1.0.1 of grand.

Ollie

[Maxim-93]

Thanks so much Ollie, I'll give that a go!

[olliemelling]

@Maxim-93 are you happy for this issue to be closed now? Happy to help further if there are still issues.

[Maxim-93]

Hi Ollie,

That sort of worked- I'm getting a new but similar error message:
110000 move(s) completed (19 accepted (0.0173 %)). Current N = 2. Average N = 2.582 Traceback (most recent call last): File "equil-uvt1.py", line 86, in <module> ghost_resids = gcmc_mover.getWaterStatusResids(0) AttributeError: 'StandardGCMCSphereSampler' object has no attribute 'getWaterStatusResids'

[olliemelling]

Hi,

The getWaterStatusResids function is only available in the more recent versions of grand, not in version 1.0.1 which is the version needed to run the scripts in this repo.

What are you using this function for? If it is just to get the resids of the ghost waters then something like this: gcmc_ids = np.where(gcmc_mover.gcmc_status == 0)[0] should work?

Happy to take a closer look if you can share the full script.

Thanks,

Ollie