No trajectory files generated with RBFE protocol (default settings)

[lwhitm1]

Describe the bug
When I run the RBFE protocol with OpenFE, the output only contains the following files:

• checkpoint.chk
• forward_reverse_convergence.png
• ligand_RMSD.png
• replica_exchange_matrix.png
• simulation.nc
• structural_analysis.npz
• db.json
• hybrid_system.pdb
• mbar_overlap_matrix.png
• replica_state_timeseries.png
• simulation_real_time_analysis.yaml
  Trajectory files (i.e., .dcd or .xtc) are not generated, though they are expected to be written during the simulations.

To Reproduce
Steps to reproduce the behavior (ideally a minimally reproducible example):
Run python local_openfe_openmm_setup.py -s inputs/final_lig.sdf -p inputs/nitazene_protein.pdb -c Nitazene_D -m 'CCN(CC)CCN1C2=C(C=C(C=C2)[N+](=O)[O-])N=C1CC3=CC=CC=C3'.
This generates the commands for submission (located in openfe_output/cmds) compatible with an HPC cluster equipped with SLURM. Transformation JSON files will be written to openfe_output/transformation.

I've linked a OneDrive folder containing:

• Input protein pdb and ligand sdf (in inputs/)
• setup script (local_openfe_openmm_setup.py)
• output folder from one of the runs (successful_job_output/)
• conda list output (env_details.txt)

Software versions

• Which operating system and version did you use? (e.g. ubuntu 22.04.5)
  OS: RHEL 8.1.0 (linux); OpenFE: 1.5.0
• Which method did you use to install this package? (e.g. conda-forge)
  Conda-forge
• Copy/paste the output of conda list (or the equivalent for your package manager):
  See env_details.txt

Output
No errors were raised. The job ran to completion. (Both .err and .out files can be found in the successful_job_output folder.)

Expected behavior
I expect trajectory files to be written when running the rbfe protocol with default settings.

Additional context
I want to confirm whether this is a bug or if trajectory output needs to be enabled by an option I've overlooked.

[IAlibay]

Trajectory files (i.e., .dcd or .xtc) are not generated, though they are expected to be written during the simulations.

Sorry just providing a very brief response for now @lwhitm1, the simulation.nc file is the trajectory file. One of us can point you to how to access it (maybe we need some docs for it).

[lwhitm1]

Hi @IAlibay! Does the.nc file need to be converted to some other filetype to be visualized? I've tried loading it into pymol, vmd, and ovito, but none of the programs can open it.

[atravitz]

ah! I think I may have led @lwhitm1 astray here (also tagging @mikemhenry as well since I think we both thought this) - looks like default .xtc writing only applies to plain MD simulations:

openfe/openfe/protocols/openmm_utils/omm_settings.py

Line 669 in d0af3f5

trajectory_write_interval: FloatQuantity['picosecond'] = 20.0 * unit.picosecond

(note: it looks like this documentation could be improved regardless)

@mikemhenry did point out that we do a conversion here: https://github.com/OpenFreeEnergy/IndustryBenchmarks2024/blob/dded1a94d15b0fcba4f806782cdd25e607d89ba4/industry_benchmarks/utils/results_cleanup.py#L121-L128

so maybe we should expose that functionality in openfe? curious what science folks think @hannahbaumann @IAlibay @jthorton

[ijpulidos]

I quickly wrote a script to help with this a while ago, it might be helpful https://gist.github.com/ijpulidos/585039061910492fc13292d1f4c5f84b

[ijpulidos]

If we end up doing a tool or API to help with this, I think it would be interesting to contribute the code upstream to openmmtools if suited

[IAlibay]

Just chiming in here quickly. There already is a tool for this, it's in openfe-analysis.

The instructions are here: https://github.com/OpenFreeEnergy/openfe_analysis/blob/main/README.md

I'll have a chat with some of the other members of the team tomorrow, maybe we can expose this in a notebook for now (whilst we do updates to openfe-analysis to improve documentation).

[IAlibay]

We'll be writing a notebook to show off how to use openfe-analysis, this should resolve this issue.