AnalysisError: SOMD free-energy analysis failed!

[velocirraptor23]

Hi,
Sorry for my delay in reply, but I have found something strange while running FEP with somd. When I launch the calculation for the RBFE using BSS.FreeEnergy.Relative.analyse, I get the follow error:
File /cluster/ddu/cmmartinez001/miniconda3/envs/biosimspace-dev/lib/python3.9/site-packages/BioSimSpace/FreeEnergy/_relative.py:605, in Relative._analyse_somd(work_dir)
598 proc = _subprocess.run(
599 _Utils.command_split(command),
600 shell=False,
601 stdout=_subprocess.PIPE,
602 stderr=_subprocess.PIPE,
603 )
604 if proc.returncode != 0:
--> 605 raise _AnalysisError("SOMD free-energy analysis failed!")
607 # Re-run without subsampling if the subsampling has resulted in less than 50 samples.
608 with open("%s/mbar.txt" % work_dir) as file:

AnalysisError: SOMD free-energy analysis failed!

I have run the same for the O-xilene tutorial and it works fine. However when I see the simfile.dat I noticed that in my calculations it does not start from step 0. I noticed in the tutorial all lambdas begin from 0 and also, the version of sire is different.

Versions:
Tutorial: 2022.3.0
My calculation: 2023.0.3

Running in RHEL 7

I suspect BSS.FreeEnergy.Relative.analyse is not able to pick the data inside because I am missing step 0, The cfg file looks pretty much the same as the tutorial setup, so, I don't think the parameters were wrongly selected.

Please, let me know if you need more details.

Best wishes,

Cesar

[lohedges]

Any chance you can try with the latest version of the code from the OpenBioSIm channel:

mamba create -n openbiosim-dev -c conda-forge -c openbiosim/label/dev biosimspace
mamba activate openbiosim-dev

You can use BioSImSpace to analyse existing data as follows:

import BioSImSpace as BSS

pmf, overlap = BSS.FreeEnergy.Relative.analyse("/path/to/your/data")

Here "/path/to/your/data" is the working directory from an existing run.

The analysis code uses a wildcard to match whatever lambda directories exist, so I don't think it should fail if a directory is missing, or has a different name.

Alternatively, if it's not too large, you could zip up your working directory and post it here. (Or share by another means if it's too big.)

Cheers.

[velocirraptor23]

simfile.txt
Hi,

Please, find attached the simfile where I think the error comes, the data inside can not be readen because it starts from 250 step and not from 0. I will check if the version you mention can make a difference.

Best wishes,

Cesar

[lohedges]

I've manually edited some local data so that all of the simfiles don't start from zero and it is still working for me. I'd try manually running analyse_freenrg to see if it gives any more information as to why it failed, e.g.:

analyse_freenrg mbar -i free/lambda_*/simfile.dat -o free/mbar.txt --overlap --subsampling

[velocirraptor23]

Hi,

It works now whith tha latest version. Thanks a lot. Perhaps something was wrong the the previous version. I used the same data.

[lohedges]

Great, I'm pleased to hear. I don't think the BioSImSpace analysis has changed in that time, but we have occasionally had to make edits to the analyse_freenrg script due to changes in the pymbar API, so maybe the version that you were using was broken.