Relative constraint deviation after LINCS

pmx
1

Hello,

I’m using the eqA.mdp file from the tutorial here:

Despite running several equilibration steps with and without restraints on my ligand I get the error:
Step 3, time 0.006 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.001067, max 0.057983 (between atoms 9731 and 9732)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length

I can see the dummy atoms swinging around like crazy when I visualize the equilibration trajectory. I don’t know why the equilibration doesn’t crash but the run started from the provided mdp file does.

Apparently an mdp setting I need to adjust?

Or is it an issue with my ligand structure? I can’t see why this would be the case since this is a structure I’ve used in many simulations (without the dummy atoms).

Thank you,

Dan

2

I switched the constraints from “all-bonds” to “h-bonds” and it eliminated the errors.

3

Switching to h-bond constraints is likely incorrect for the pmx based ligand hybrid topologies. In case, a hydrogen atom is mapped to a dummy in stateA, the constraint will not be generated for the bonds involving this dummy.

I would suggest trying an SD integrator (note that you need to change tau-t in this case).

1 Like
4

Okay, thank you. In the case of an SD integrator, I would return constraints = all-bonds?

Thanks, Vytas.

Dan

5

Yes. It is not guaranteed that the SD integrator will help, as there might be other issues, but it is worth to try it out.

Is this simulation started right after energy minimization? If so, what is the maximal force at the end of the EM?

6

The first time around I ran nvt and npt equilibration.

This time I went straight from em to the included eq mdp. The maximum force was 2.203x10^3.

Could I go from the eq run with md integrator and my h-bond constraint to the TI runs with the sd integrator all-bond constraint without issue?

7

I would recommend to retain consistent use of constraints and integrator for both steps: eq and transitions.

1 Like
8

Encountering similar error(Relative constraint deviation after LINCS) at the last lambda window( at lambda=1) and is for the atoms which are transformed to dummy. I have been using h-bond constraints and not all-bond constraints cause it does not involve any hydrogen to heavy atom morphing. Would using SD integrator alleviate the issue? Or it has to do with free energy settings (note that it is a K2A transformation)?

9

could it be that you are using ‘master’ branch of pmx? In that case, you need to use all-bond constraints. h-bond constraints are supported on with the ‘develop’ branch

10

I am using the develop branch. I am attaching the topology for the same. Using all-bonds must solve the problem cause the dummy atoms are causing the problem. Is there any other issue which might have been causing it?

~/Softwares/pmx$ git branch 
* develop
  master

PMX_K2A
PMX_K2A1311×578 95.2 KB

11

Compared the same with tripeptide data as provided in the website, it is the same topology.

12

Webserver is still using ‘master’ branch, this means that you are likely also using ‘master’ branch, thus you need to use all-bond constraints. Maybe your GMXLIB is still pointing to the force field folder in the ‘master’ branch? For h-bond constraints, switch to ‘develop’ branch. Make sure to use amber99sb-star-ildn-mut.ff, amber14sbmut.ff or charmm36m-mut.ff force field.

1 Like
13

yes, got that. This was the reason for the problem. Thanks a lot.

14

Would it differ in deldelG values with the use of different constraints

15

Both legs of the thermodynamic cycle need to use the same simulation parameters, i.e. the same constraints

1 Like
16

Certainly, thanks.