Possible inconsistency when mixing GROMACS 2023.1 and 2025.x for umbrella sampling workflows

User discussions
1

GROMACS version: 2023.1
GROMACS modification: No
Hello everyone,

I would like to ask about best practice regarding GROMACS version consistency.

I am running umbrella sampling simulations, and part of my workflow was originally prepared/tested with GROMACS 2023.1. Later, on an HPC system, the available GROMACS module was 2025.x, so the current production runs are being generated and executed with GROMACS 2025.x.

My question is:

Is there any risk of inconsistency if earlier preparation or test runs were done with GROMACS 2023.1, while the final umbrella sampling production runs are generated and run with GROMACS 2025.x?

More specifically:

  1. Should all .tpr files be regenerated with the same GROMACS version used for mdrun?
  2. Is it safe to continue from .cpt files if the GROMACS version changes between 2023.1 and 2025.x?
  3. For publication/reproducibility purposes, is it acceptable if all final production logs are from GROMACS 2025.x, even if some earlier testing/preparation was done with 2023.1?
  4. Are there known differences between 2023.1 and 2025.x that could affect umbrella sampling results or checkpoint continuation?

I want to make sure that my final dataset is version-consistent and scientifically reproducible.

Thank you.

2

GROMACS should always produce the same (not binary, but in terms of ensembles) results, unless there are bugs that have been fixed or new bugs introduced.

3

Thank you very much for your clarification. This is very helpful and reassures me that using GROMACS 2025.x for the final production runs should not introduce systematic inconsistencies, provided that no relevant bugs are involved.

I will ensure that the final production workflow and the GROMACS version used are clearly documented for reproducibility.

Thank you again for your time.