AWH Error During Ligand-Membrane Pulling in Z-direction

User discussions
,
1

GROMACS version: 2025.0
stepX_pull3.mdp (3.2 KB)

GROMACS modification: Yes/No

Dear All,

While attempting to run an AWH-based pulling simulation, I’m encountering an issue where a compound is biased to move through a lipid membrane along the Z-axis.

I am using GROMACS version 2025.0, and my simulation is based on an AMBER force field setup. The goal is to observe the compound’s passage from one aqueous region through the membrane and into another aqueous region.

The simulation setup is as follows:

==Edge
Water-1
Compound (initial position)
Water-1
Membrane (~40 Å thick)
Water-2 (~32 Å thick)
==Edge
Water region (1) has a thickness of ~30–32 Å. The compound starts in the middle of Water-1 and should pass through the membrane to reach Water-2. The pulling is applied in the +Z direction using AWH.

However, the simulation crashes early on with the following error:

#======
starting mdrun ‘Title’
50000 steps, 100.0 ps.
step 400, remaining wall clock time: 29 s

Program: gmx mdrun, version 2025.0
Source file: src/gromacs/applied_forces/awh/biasstate.cpp (line 1743)
Function: gmx::BiasState::getSharedPointCorrelationIntegral(int) const::<lambda()>

Assertion failed:
Condition: sharedCorrelationTensorTimeIntegral_[gridPointIndex][i] == 0
Correlation tensor time integral of unvisited points should be 0.
#======

I’ve attached the .mdp file used (stepX_pull3.mdp). Pull groups and potential provider appear to be configured correctly. The compound’s COM is within the AWH sampling range at the start of the run.

  1. Could you please help identify what might be triggering this assertion failure or how best to initialize the system or AWH state to avoid it?
  2. How to make sure the compounds pass the membrane (and reach near to the middle of Water-2)

Thanks!

2

The user should never get an assertion failure. But this one is hard for us to avoid.

There are several issues in your setup:

  • You seem to think that the pull coordinate is the absolute z. But it is relative to center of bilayer, which has coordinate 0.
  • With a membrane, you should use pull-geometry=cylinder.
  • Your diffusion constant looks far to high. But this is likely compensated by the small initial error.