normal Cholesterol problem with newer GROMACs version

  • kelley
  • kelley's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
1 year 1 month ago #9607 by kelley
Hello Martini group,

I'm having an issue with a new GROMACs version that was updated on my local supercomputer. I had run a mixture of phospholipids, sphingomyelin and cholesterol (V2 with virtual sites) just fine with a prior version of GROMACs but after the supercomputer's update, now I'm getting LINCS WARNINGS for cholesterol in my Martini 2 production run. Here is a view of one of the errors:

Step 259450, time 5189 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.487373, max 4.677094 (between atoms 10895 and 10894)

Step 259450, time 5189 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.724989, max 7.359502 (between atoms 4031 and 4030)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
11191 11190 44.9 0.2719 0.2714 0.2719
4034 4031 95.9 0.5376 1.6339 0.5376
4034 4030 89.8 0.7238 0.8307 0.7237
4031 4030 93.4 0.2719 2.2729 0.2719
4030 4028 73.4 0.4905 2.1186 0.4904
2751 2750 48.8 0.2718 0.2784 0.2719


I've attempted to relax my LINCS constraints, ran longer equilibration with even smaller time steps, with similar results every time in production runs. To make sure it was cholesterol, I ran a simulation without cholesterol (just phospholipids and sphingomyelin) and obtained no errors/warnings in my production run.

I've come across this recent paper ( https://doi.org/10.1021/acs.jctc.2c01032 ), which mentions the issue with significant temperature gradients between cholesterol and phospholipids which they provide an itp of the optimized cholesterol version. I haven't tried this new topology file, however I'm still wondering if anyone else has run into this issue or knows why an updated GROMACs would crash cholesterol when it worked before.

I appreciate any insights anyone could provide.

Please Log in or Create an account to join the conversation.

More
1 year 1 month ago #9610 by MatsMD
I had some strange LINCS errors too for CHARMM36m lipids (gromacs-2020, forgot exact build) which did not occur in gromacs-2022.3. Have had no issues with Martini 2 lipids.

My search then led me to this thread: gromacs.bioexcel.eu/t/lincs-error-on-new...ing-gmx2020-3/828/12 , where they do mention the issue seems to be fixed in 2020.4, so it's still odd you see this.

Not an expert but the issue seems to lie at the compiling stage and CUDA libraries. Perhaps the different versions on your cluster are compiled differently / use outdated CUDA libraries? You could try compiling the problematic GMX version on a local machine to see if the problem persists. Any other systems that are more unstable in the newer GMX version?

Please Log in or Create an account to join the conversation.

Time to create page: 0.085 seconds