Some ABFE questions/observations

[sboresch]

Hi all,

I have currently a visiting student, Julian ( @JulianSierraRodriguez ) who is trying out OpenFE's ABFE functionality. We have some questions/observations and would appreciate your feedback:

(1) Julian followed the T4 lys./ toluene tutorial example; here are the values he obtained

OpenFE 1.9.1:

-3.14 kcal/mol
-4.16 kcal/mol
-4.26 kcal/mol
OpenFE 1.8: (a first attempt he already tried in Barcelona prior to coming to Vienna)

-4.61 kcal/mol

In terms of simulation (equilibration/production) lengths, we used the tutorial values. Do these numbers sound about right? They look OKish to me, and we also have results for two other ligands from this series and both are in the right direction compared to experiment.

(2) a nit-pick: OpenFE 1.8 seems to report directly the results of the protein-complex leg, the solvation leg, and the restraint correction. In OpenFE 1.9, most of this info seems hidden, and we had to turn on logging to get some relevant info. However, even with logging on, we don't see the restraint (standard state) correction contribution explicitly. Probably running with logging at the 'info' level is a good idea anyways, but the OpenFE 1.9.1 output seems too terse.

(3) We are now starting some simulations with p38. Unless I am completely mistaken, you showed ABFE results for this system in the previous to last TAC meeting, but only as a scatter plot compared to experiment. Could you dig out by any chance the actual numbers? This would help us further to ascertain that our installation works correctly etc.; in particular, as Julian would then move back calculations to their 'big(ger)' resources in Barcelona rather than hogging GPUs on my humble local cluster.

[IAlibay]

Hi @sboresch,

Thanks for opening this discussion.

(2) a nit-pick: OpenFE 1.8 seems to report directly the results of the protein-complex leg, the solvation leg, and the restraint correction. In OpenFE 1.9, most of this info seems hidden, and we had to turn on logging to get some relevant info. However, even with logging on, we don't see the restraint (standard state) correction contribution explicitly. Probably running with logging at the 'info' level is a good idea anyways, but the OpenFE 1.9.1 output seems too terse.

I am specifically triaging this (others will chime in for your other questions).
Could you provide more information on what you were doing to get that information in that past that isn't working now?

Is this by calling openfe gather-abfe --report raw path_to_results_jsons or some other output?

[JulianSierraRodriguez]

Hi @IAlibay ,

We are using a python script with the code of the tutorial (https://docs.openfree.energy/en/latest/tutorials/abfe_tutorial.html) where we use the python API and not the CLI.
Before when analyzing the results in the json file (https://docs.openfree.energy/en/latest/tutorials/abfe_analysis_tutorial.html), we got the overall energy, complex and solvation legs energies but also the standard state correction, this last term seems to be missing now.

[IAlibay]

Thanks for getting back so quickly @JulianSierraRodriguez. I think I know what might have happened there - we might have forgotten to update our docs since the underlying API changed a bit, sorry about that!

[sboresch]

So, what should we have changed/adapted?

[IAlibay]

For now, I would suggest using the CLI's openfe gather-abfe command to get outputs from the results JSON files, https://docs.openfree.energy/en/latest/reference/cli/gather.html#openfe-gather-abfe

In v1.9+ that should give you a breakdown of the results when using the "raw" report.

P.S. There are some slight issues with the reporting, so if you can wait until the v1.10 release of openfe (hopefully early next week) that will work better.

[JulianSierraRodriguez]

With the CLI's openfe gather-abfe it seems to give us those corrections, thank you.