Lateral Stress profile from membrane simulation

User discussions
1

GROMACS version:2018.6
GROMACS modification: No
Hello all,
I want to plot lateral stress map from MD simulation of lipid bilayer i.e I want to plot local stress of bilayer as a function of x and y. Is there any Gromacs module or external tool that I can use?
* I have tried Gromacs-LS (a custom version of GROMACS developed for local stress calculations from molecular simulations), but it seems that it is based on Gromacs 4.6.6 and can’t handle files generated from later versions (2018.6 in my case).
Thank you.

2

Hi -

For Gromacs-LS you need to regenerate the TPR using gromacs-LS and your mdp/starting structure/topology - you may need to tweak the mdp as there are some changes between versions. You can then do the analysis using the new tpr and the .trr from the original run.

3

Hello @kevinboyd,
Thanks for your suggestion, it works very well!
Though I have a one query. There is 32000 atoms and 500 frames in the trr file. As I see, it is taking very long time to analyze each frame. So, is this normal of the software? This may sound stupid because it depends on hardware and so on. But can you shed any light on this from your experience? It would be really helpful.
-Anirban

4

Not stupid at all, but I think most of the information on this is in the Gromacs-LS manual if you give it a close read, including the FAQ section -

GROMACS-LS is very slow because it doesn’t use most of the optimizations that Gromacs has built in (no SIMD acceleration on CPUs, no GPUs) and it runs on only one core. Another reason is that there’s no PME for electrostatics, so you have to use a larger short-range electrostatic cutoff to get accurate forces - have your read about that parameter in the Gromacs-LS doc? It’s pretty important to get right.

The good news is that if you have a cluster or computer with multiple cores, you can break up that trr file into chunks (Gromacs-LS provides a tool for this), use a bash script to call the program on each chunk (one for each cpu core you have is probably best), and then combine the results at the end during the processing step.

5

Thank you so much @kevinboyd . I will look into the software documentations more carefully and of course try to use scripting to speed up the calculation. Thanks a lot!
-Anirban