Use cases

[MichaelClerx]

For the final product

What we know so far and open questions

[MichaelClerx]

1a. High-throughput QC and export

• Load data and meta data for lots of wells from syncropatch or similar format
• Post-process
• Run QC
• Export data and meta data for selected wells
  • How will we store data? CSV preferred for portability, binary formats for size and meta data
  • How will we store meta data (Rseries, Cm, gLeak, Eleak, protocol, ramps, etc) CSVW, contained in binary format
• Generate report
• Generate data for interactive viewer showing which wells passed which criteria, what would happen if thresholds are changed etc

[RomanSyunyaev]

I am relatively happy with what we have in subtraction_qc.csv at the moment.
A csv file combining

1. metadata
2. all the parameters determined by pcpostprocess (like Erev\Eleak\gleak from the ramps).
3. QC values (mostly True\False).

So something similar should be outputed, IMO.
Another option would be to have separate files for 1-3, which might be more tidy, but less convinent.

[MichaelClerx]

1b. Optimised variant?

• Load meta data (may need changes in syncropatch_export)
• Early reject based on meta data QC
• Load data for passing cells
• Post-process
• Run QC
• Export data and meta data for selected wells
• Generate report
• Probably not interactive viewer? Unless we add in some minor complications e.g. "Pass/Fail/Early-fail". So maybe green/red/gray.

[MichaelClerx]

Since meta data (Rseries, Cm, Rleak) criteria are quite forgiving, perhaps not being able to change their cut-off criteria in the online viewer isn't so bad, and 1b is actually the main use case?

[MichaelClerx]

1c Partial-plate variant

• As before, but needs changes in syncropatch_export to stop loading all data (np.memmap instead of np.fromfile)
• Some way of selecting rows/columns instead of listing wells would be nice-to-have

[MichaelClerx]

2a. Manual data, external meta data, and export

• Loop over list of file/folder names
• Load meta data from? A csv file? A list written directly into the script?
• Load data (ABF) w Myokit
• May need to split file into multiple parts, e.g. staircase, other protos, staircase
• continue as in case 1

[MichaelClerx]

2b. Manual data, mostly internal meta data, and export

• Loop over list of file/folder names
• Load meta data from HEKA file w Myokit
• Still need to add manual data. Protocol and ramps are in format, but things like "which one is the leak step" can't. Also exact alpha/beta for Rs compensation
• Load data
• continue as in case 1

[MichaelClerx]

3a. Partially on the fly

Part 1

• As in 1 or 2, but don't export data, just list of passed wells/cells
• Includes report etc.

Part 2

• Load from original format, redo post-processing
• Fit, plot, or whatever you want to do

[MichaelClerx]

3b. Fully on the fly

Looping over cells:

• Load meta, early reject
• Load data for passed, post-process
• QC: move to next cell if fails
• Fit, plot, etc.

[MichaelClerx]

Downside would be if this crashes in fit 10 out of 50 you'd have to have a clever way of restarting without losing the first 9 fits. Also post-processing has to be completely non-stochastic