Bug #2767

grompp warns about md + FEP even for fully coupled states

Added by Mark Abraham about 1 year ago. Updated 2 days ago.

Fix uploaded
preprocessing (pdb2gmx,grompp)
Target version:
Affected version - extra info:
probably all versions since about 5.0
Affected version:


gmx grompp with FEP .mdp inputs and non-SD integrators issues

WARNING 1 [file grompp.mdp, line 62]:
  For proper sampling of the (nearly) decoupled state, stochastic dynamics
  should be used

It does so even for a fully coupled state. We should not be issuing warnings when there's no data to support that, because that will train users to use -maxwarn.

We should check if this warning is still appropriate
  • at least check the git history for which Redmine added it - is full VDW but no charge OK?
  • is the warning based on ancient implementations and needs updating?

If we keep it, it should be sensitive to the calculation being done. For example, even slow growth (TI) starting from a not-nearly decoupled state should not issue it. FEP should only issue it for suitable values of lambda (if any).

fep-warning-repro.tgz (5.88 MB) fep-warning-repro.tgz tpr and grompp inputs Mark Abraham, 11/19/2018 05:34 AM


#2 Updated by Berk Hess about 1 year ago

The user can manually modify the topology to decouple a molecule. Do we want to check for that? Or do we rather want to reduce false positives and accept some false negatives? Then we would only need to check if we have a fully decoupled state, i.e. couple-lam0/1 = none with lambda=0/1.

#3 Updated by Berk Hess 2 days ago

  • Status changed from New to Fix uploaded
  • Assignee set to Berk Hess
  • Target version changed from 2020 to 2019.5

Uploaded a "fix" that changes the warning to a note.

Also available in: Atom PDF