Add tool to extract equilibrium pose from simulations

[JenkeScheen]

It would be super useful for debugging and explaining perses ddG predictions if we would be able to write ligand poses that represent equilibrium during the simulations. This could e.g. be in the form of a PDB snapshot every n steps, where n is a user-set flag?

[ijpulidos]

@JenkeScheen Thanks for opening the issue!

Just trying to clearly understand this. We are currently working on a streamlined way of extracting the trajectories from the simulation results (.nc files). I guess we could make it in a way that we extract every n steps and also initializing it after we have detected equilibration or something like that. Would this satisfy your requirements?

[JenkeScheen]

Thanks @ijpulidos, it would. The use-case here is that when I present FE predictions to chemists that are trying to explore a sub-pocket with a ligand series they want to be able to do some sort of qualitative assessment that the R groups they came up with are hanging out in the intended areas (and that the ligand isn't just jumping out of the pocket, of course). A collection of snapshots (preferably at ~ equilibrium) would allow that assessment.

Ultimately we would want to be able to automate generation of a GIF showing a simulation of the ligand in the pocket but snapshots are a great starting point!

[jchodera]

I'll plan to implement both a CLI and Python API to extract equilibrium snapshots of both initial and final molecules.

For a CLI, how does this look as a first attempt?

perses extract-snapshots --nsnapshots 10 --initial initial.pdb --final final.pdb

Ultimately we would want to be able to automate generation of a GIF showing a simulation of the ligand in the pocket but snapshots are a great starting point!

For the record, this is a nice thing to have, but it will never be as useful as being able to interactively inspect the structure (e.g. in fragalysis). But I can see utility if we can generate both the GIF of 3D interactions and a 2D interaction diagram and push this to the real-time dashboard.

[JenkeScheen]

For a CLI, how does this look as a first attempt?

Input:

1. Are .pdb files enough as input or do we need to inject e.g. .nc as well?

Output:

1. Perhaps an --out flag to allow custom out names?
2. Would the output be a multi-pdb of nshapshots snapshots of the PL complex?
3. It could be valuable to generate a large number (10-100) of ligand poses, perhaps as a multi-sdf?

For the record, this is a nice thing to have, but it will never be as useful as being able to interactively inspect [..]

Agreed, interactive is king. Although, beside being a decent placeholder until we can upload this type of visualisation to e.g. fragalysis, GIFs are nice because they can be easily embedded in slidedecks without weighing them down too much.