Description
I would like to analyze free energy simulations ran with openMM using the openmmtools toolkit.
This issue is related to #296 and choderalab/openmmtools#652 .
There is some code already available at #296 (reply in thread) from @xiki-tempula which looks good to me and works for me on a Hamiltonian replica exchange simulation with a single lambda variable. I have tested this using alchemlyb 1.x with pymbar 3.x as well as alchemlyb 2.4.1 and pymbar 4.0.3 and it seems to produce plausible results that agree with the estimates provided by openmmtools.
The main issue that came up in the discussion above was that the lambda values are not easily accessible from the openmmtools output. For single-lambda-variable simulations using FEP-based estimators this should not be necessary though. Openmmtools also does not save dU/dL values (at least not that I am aware), so an implementation that only supports FEP-methods should be sufficient.
I also don't know what a simulation with multiple lambda variables looks like in the output file, but it might be enough to support only simulations with one lambda-variable.