Local stress first moments for Martini 3 cholesterol: constraints and type-4 virtual sites

User discussions
1

Hello,

I am looking for advice on a suitable workflow for calculating the first moment of the lateral stress profile in Martini 3 lipid bilayers.

My systems are POPC bilayers with and without 30 mol% cholesterol, simulated using standard GROMACS 2025.4. My eventual aim is to estimate monolayer spontaneous curvature from the stress-profile first moment together with an appropriate monolayer bending modulus.

I am considering GROMACS-LS. I understand that it is a modified version of GROMACS rather than part of the standard distribution, so advice on an appropriate implementation or a more suitable support channel would also be appreciated.

The Martini 3 cholesterol topology I am using is from:

Borges-Araújo et al., “Martini 3 Coarse-Grained Force Field for Cholesterol” (2023), DOI: 10.1021/acs.jctc.3c00547.

It contains three explicit distance constraints and five virtual_sites3 function-type 4 sites.

My main concern is whether the contributions from these constraints and virtual sites are correctly represented in the spatially resolved stress profile, particularly its first moment. Correct reproduction of the total virial alone would not resolve this concern.

Could anyone advise on the following aspects of this workflow?

Is there a GROMACS-LS version, or another implementation, known to support this combination of constraints and virtual sites for local stress calculations?

Which force-decomposition settings and validation checks would be appropriate for the stress-profile first moment?

Can the required analysis be performed using existing XTC trajectories, or would additional simulations with velocities saved to TRR be necessary?

I have the XTC trajectories, topology and simulation input files, and final checkpoints, but no production TRR trajectories containing velocities. I can run additional simulations if needed.

This is a workflow-planning question; I have not yet generated local stress profiles. I have also contacted the GROMACS-LS developer by email and am awaiting a response.

Thank you for any guidance or relevant references.

Best regards

GROMACS version:2025.4 (Ubuntu build: 2025.4-Ubuntu_2025.4_1)
GROMACS modification: No

2

Hi,

I tried to use GROMACS-LS some years ago also on lipid:cholesterol bilayers simulated with the MARTINI force field. Unfortunately, it did not work out due to the virtual sites. If I remember correctly the problem was how the virtual sites distribute their force on the constructing atoms. There is also an article by Markus Deserno briefly mentioning this problem (DOI: 10.1016/j.bpj.2019.11.3398).

If you want to use GROMACS-LS you need a .trr file with the velocities and the forces saved at a high output frequency. However, you can generate the trajectories with a new GROMACS version and then use GROMACS-LS for post-processing. There is an old tool (2009!) linked on the MARTINI website ( Legacy Tools – Martini Force Field Initiative ). I have never used it and cannot comment on the method, but maybe you want to take a look.

Best,
Marius

3

Hi Marius,

Thank you for sharing your experience and the reference. It is helpful to know that you encountered the same issue with virtual sites in Martini lipid–cholesterol bilayers.

For now, I have decided to put local stress and spontaneous curvature calculations on hold and focus on structural and dynamical analyses of my membrane simulations.

Thank you as well for explaining the post-processing workflow and pointing me toward the legacy tool. I will keep these suggestions in mind when I revisit local stress calculations.

Best regards