Nans in an unstable Non-eq. cycling calculation (across a mutation)

Hi!

I’ve been setting up a few mutants to study the impact of a mutation in the context of a dimerization, and I have encountered a mutation case that appears to fail? It is an ILE->ASP mutation.

I’ve attached a tarball here of the setup files + the scripts used. ripk2-failure-case.tar.gz

The three python scripts are run as shown in the run_scripts.sh. For both Apo and Complex phase, the HTF setup seems to work(setup_mutation_htf.py), or at least ends with no error reportings/warnings. However, upon trying to do non-equilibrium cycling (run_neq.py) it instantly fails with the following stack trace within the first step (also documented in log.failure):

WARNING:root:Warning: importing 'simtk.openmm' is deprecated.  Import 'openmm' instead.
DEBUG:root:Step: 0, Equilibrating at lambda = 0, Elapsed time: 0.020032151540120443 minutes
Traceback (most recent call last):
  File "/lila/data/chodera/sukrit/ripk2/perses/lilac-setup/i307d/scripts/run_neq_complex.py", line 129, in <module>
    integrator.step(1)
  File "/home/singhs15/miniconda3/envs/dev-perses/lib/python3.10/site-packages/openmm/openmm.py", line 6011, in step
    return _openmm.CustomIntegrator_step(self, steps)
openmm.OpenMMException: Particle coordinate is nan

I have used this exact scripts for three other mutations on this system: SER->ASP, GLU->LYS (in 2 different GLU positions) all of which have run successfully 1 cycle to completion. In another protein system I have been able to use these scripts to do a neutral->charged residue mutation (GLY->ARG) without fail so I’m not too sure what the issue might be here?

After chatting with @ijpulidos I tried running the minimization with larger tolerance (increased by 10X) or using the FIREMinimizationIntegrator and both cases showed the exact same behavior for me.

Any advice as to what to do? I took a look at the start and end PDB files and everything seemed okay for me there too so I’m not too sure how to progress.

[sukritsingh]

What are you setting maxIterations equal to? Try setting it to 0 if it isn’t already set to that. From the openmm docs:

I used the default maxIterations=0 based on the documentation for LocalEnergyMinimizer for 7.7

[ijpulidos]

Thanks for the feedback @sukritsingh

I can reproduce your issue. I will have to take a deeper look into it to see if I can spot something. At a first glance everything seems okay.