Too many LINCS warnings with Comm-mode Angular

User discussions
1

GROMACS version: 2019.1
GROMACS modification: No
Here post your question

I am conducting simulations of a peptide in gas phase using opls forcefield. Box size ~ 999 nm^3 and cutoffs 333.3 nm. Although I understand that angular is the correct comm-mode to be used here, I get too many LINCS warnings when I set comm-mode to angular but none when linear. Any thoughts?

2

Hi Neena,

I would guess here that large movements are introduced when using the angular comm-mode due to rotating a large box. If you must use comm-mode angular, a first shot would be to equilibrate the system with comm-mode linear for some steps, then start the simulation with comm-mode angular.

3

This should work when the molecule is not crossing box edges. Where did your place your molecule in the box?

4

@cblau That did not work, but thank you

5

@hess using gmx editconf command, placed in the center of box -c -d 496.5 -bt cubic

6

Could you attach your tpr and md.log files so I can have look?

7

@hess is there any plan to re-implement proper non-periodic vacuum calculations? The workarounds of large boxes seem to have lots of unexpected pitfalls. I know there is an open issue on this (https://gitlab.com/gromacs/gromacs/-/issues/3526) but wanted to see if there is any momentum as the request has sat dormant for a few months.

8

@hess The forum does not allow to upload .tpr or .log files, instead I am attaching here my most recent md.mdp file where I just changed from angular to linear comm-mode to circumvent lincs warnings with the former mode. md.mdp (2.0 KB)

9

I realized there could be an issue if your peptide undergoes large conformational transitions, as then the orientational fit might suddently change the orientation of the molecule. Do you have large conformational transitions?

10

I plan to re-implement non-periodic boundary conditions since a long time, but this has still not happened because other things have had higher priority. The main working is making sure the domain decomposition works properly and efficiently. But I just realized that maybe supporting this without domain decomposition would cover most use cases.

11

I did not try with angular mode as I did not want to override lincs warnings. But yes I see significant conformational transitions with linear mode. Please see this post where I show those transitions Rmsd shots after recentering with trjconv