Add updated tool to consolidate and preprocess data from simulation results

[mcwitt]

Starting from run description in input JSON file:

    "1": {
        "JOBID": 1,
        "directory": "RUN0",
        "end": 0,
        "end_pIC50": "5.44251008457101",
        "end_smiles": "Cc1ccncc1NC(=O)Cc2cc(cc(c2)Cl)O[C@@H]3CC(=O)N3",
        "end_title": "TRY-UNI-2eddb1ff-7",
        "ff": "openff-1.2.0",
        "ligand": "nucleophilic_displacement_enumeration_for_FEP-sorted-x10789.sdf",
        "protein": "../receptors/monomer/Mpro-x2646_0_bound-protein-thiolate.pdb",
        "start": 2,
        "start_smiles": "Cc1ccncc1NC(=O)Cc2cc(cc(c2)Cl)OCS(=O)(=O)N[C@H]3C[C@H]4C[C@@H](C3)[NH2+]C4",
        "start_title": "EN300-784608",
        "target": "SARS-CoV-2 Mpro"
    },

1. Extract work values from simulation data (globals.csv). See example on server at /home/server/server2/projects/available/covid-moonshot/consolidate-work-to-pandas-13420.py. Create work.json files with extracted data (at run/clone/gen level).

2. Map free energy analysis over run/clone/gens. Produce the following values

   • delta_f
   • ddelta_f (BAR error estimate)
   • delta_f_low
   • delta_f_high
   • bar_overlap
   • number of samples (if we eliminate forward work, should we eliminate corresponding reverse?)

3. Collect results into an augmented version of the input JSON, with the following additional schema:

   "1": {
       "complex_phase": {
             "delta_f" : -2.345,
             "ddelta_f" : 0.012,
             "delta_f_low" : -2.567, 
             "delta_f_high" : -2.1,
             "delta_f_bootstrap" : [-2.452, -2.23],
             "nwork" : 105,
             "bar_overlap" : 0.984,
          },
       "solvent_phase" : {
             "delta_f" : -2.345,
             "ddelta_f" : 0.012,
             "delta_f_low" : -2.567, 
             "delta_f_high" : -2.1,
             "nwork" : 105,
             "bar_overlap" : 0.984,
          },
       "binding" : {
             "delta_f" : -2.345,
             "ddelta_f" : 0.024,
       }
   },

[jchodera]

These fields are optional, and can come later:

• delta_f_low, delta_f_high are 95% CI for bootstrapping
• delta_f_bootstrap can hold bootstrap replicates if we add bootstrapping later
• bar_overlap can hold the BAR overlap (0 = bad, 1 = good) once we figure this out, as a separate measure of quality control

I propose we also add core_rmsd as an optional metric to complex_phase where we can later add a step that computes the RMSD of the core atoms after protein alignment. A low number (e..g < 2.0A) would indicate the scaffold interactions are not disrupted, and can be used as additional quality control.

We can break these optional steps into separate issues to remind us to add these features later as quality control measures:

• bootstrapping BAR uncertainties
• BAR overlap
• core RMSD