Closed jchodera closed 8 years ago
These are referring to nonbonded exceptions? If so, I'm not sure it makes a difference in the current setup (since GeometryEngine
doesn't use nonbonded terms). Or am I not understanding what that warning means?
I'm not sure. Maybe we should just report this to the OpenMM issue tracker, since it sounds like a bug---we shouldn't be getting that warning if we don't have NonbondedForce
terms.
But I think the System
that is fed into parmed
there actually does have NonbondedForce
terms.
Basically, the GeometryEngine
and friends make a Structure
out of the full system (this does include nonbonded forces), and also create a new system that only has valence terms. However, I don't believe parmed
ever sees the latter system.
OK, I had misunderstood the problem. It must be happening here, right? That means that our SystemBuilder
is generating systems with incomplete exceptions.
I wonder if this is an issue with the small molecules built using the residue templates from gaffTemplateGenerator
. If so, we should see these warnings with the kinase inhibitor test system but not peptide test systems.
Actually, we could just make a simple test script that parameterizes some small molecules and creates a Structure
from them. Let me test that.
Ok. What is an incomplete exception, by the way? Is it not properly specified? Shouldn't that throw an, er, Exception
?
Oh, maybe it's that parmed
thinks there should be an exception there, but there isn't?
Oh, maybe it's that parmed thinks there should be an exception there, but there isn't?
I think that's it. Let me see if parmed complains about imatinib alone.
I wonder if this warning is due to the same reason that OpenMM adds (i,j,k,i) torsions since parmed
uses the same kind of defective algorithm that doesn't check if first and last atoms are the same:
https://github.com/ParmEd/ParmEd/blob/master/parmed/openmm/topsystem.py#L392-L398
Added a parmed fix here: https://github.com/ParmEd/ParmEd/pull/625
I think this is fully resolved with those two PRs.
Just saw this for the first time: