remove AbsoluteMeasurement concept

[richardjgowers]

introduce GroundState object

instead all measurements are relative to a GroundState

Description

Provide a brief description of the PR's purpose here.

Todos

Notable points that this PR has either accomplished or will accomplish.

• TODO 1

Questions

• Question1

Status

• Ready to go

[codecov]

Codecov Report

Merging #84 (43d78c1) into main (69ed025) will increase coverage by 2.76%.
Report is 8 commits behind head on main.
The diff coverage is 96.61%.

Additional details and impacted files

[IAlibay]

Overall lgtm, I really like this idea.

Only comments are:

1. A bit more documentation on behaviour would be good
2. It's hard to tell if this is good because I like it, I'd be ok with merging this quickly, but if we can have someone else gloss over the idea afterwards that'd be grand.

[IAlibay]

worth documenting that ground state is assumed to be the same for all inputs for this method?

[IAlibay]

Also is there a unique string we could give it so it hashes differently?

[IAlibay]

Could you introduce the idea of label here so that folks can understand that it can be set to different states that hash differently?

[jchodera]

Is "Ground state" a reference state with absolute free energy defined to be zero? If so, maybe ZeroState or ReferenceState is more clear?

DiffNet permits multiple absolute free energy measurements of multiple states relative to that zero. Will it still be possible to have multiple absolute measurements?

[richardjgowers]

@jchodera GroundState is a reference state, TrueGround is the theoretical zero point, and is a special case of GroundState. "Absolute" measurements are against one of these two types of points.

For example, MLE would first place computed ligands against an arbitrary MLE GroundState, then next draws a measurement between that GroundState and the GroundState of the experimental (absolute) measurements. The derived absolute computational values are then the sum of the edge to the arbitrary ground and the edge to the ground of the experimental points.

flowchart TD
  G[Experimental GroundState] -- experiment --> A[experimental A];
  G  -- experiment --> D[experimental B]; 
  H[MLE GroundState] --> B[computational A];
  H --> C[computational B];
  B -- RBFE --> C;
  G -- derived --> H;
  Z[True Ground] -- ??? --> G;

Loading

[IAlibay]

Is there really a need for something like TrueGround? If I'm being only slightly cynical (more likely just a realist), the standard deviation around even high quality experiment is such that you are really just setting your ground state to a given set of experimental conditions at a given time in space.

edit: I think my main issue is with the word True maybe just ReferenceGround would work? Then folks could assign a reference ground state, but not necessarily assert that this is the sole "truth".

[richardjgowers]

this diagram is better actually:

flowchart TD
  G[Experimental GroundState] -- experiment --> A[ligand A];
  G  -- experiment --> B[ligand B]; 
  H[MLE GroundState] -- MLE --> A;
  H -- MLE --> B;
  A -- RBFE --> B;
  G -- derived --> H;
  Z[True Ground] -- assumed 0 --> G;

Loading

@IAlibay TrueGround is maybe unnecessary? The way some sort of FEMap.get_absolute('ligand A') would work would be to pathfind from TrueGround to ligand A, which would be the sum of three edges (one of which is zero, so could collapse it away?).