[documentation] candidate processing suite examples

[neuromusic]

To help the most users understand how to get their data into NWB, I propose that we target a handful of dominant preprocessing pipelines (that is, spike sorting & calcium segmentation/ trace extraction) & give examples of getting that data into NWB.

• One of the pain points of getting data into NWB is accounting for provenance information related to a processing pipeline (i.e. saving the ROI and neuropil subtraction in addition to the calcium traces)
• While we won't promise a script that does a conversion, this should help get users that use these existing processing pipelines 90% of the way there.
• Each of these pieces of software outputs a fairly standardized set of files.

I'd like to brainstorm and identify a handful of target "sources" for ingest into NWB.

We can then either write examples ourselves or solicit the developers of the respective tools to contribute.

My initial brainstorm:

kilosort / phy TemplateGUI
This is the dominant spike sorting software for dense ephys recordings.

• modality: ephys
• output files: npy & tsv
• docs on outputs: http://phy-contrib.readthedocs.io/en/latest/template-gui/#analysis
• developers: @kwikteam

suite2p
AFAIK, this is the dominant segmentation and trace extraction processing software around right now.

• modality: 2P
• output files: mat
• docs on outputs: https://github.com/cortex-lab/Suite2P#iii-outputs
• developers: @marius10p

[oruebel]

I think that's a good idea. The more usable examples we have the better. For spike sorting other candidates may be: 1) KlustaKwik http://klusta.readthedocs.io/en/latest/ and 2) MountainSort. @neuromusic do you already have someone in mind that could help with putting these examples together?

[neuromusic]

klustakwik is a good candidate, too... that would be the KwikGUI version of phy.

klusta(kwik) / phy KwikGUI
This is the dominant spike sorting software for <100ch silicon ephys recordings.

• modality: ephys
• output files: Kwik format (HDF5)
• docs on outputs: https://github.com/klusta-team/kwiklib/wiki/Kwik-format
• developers: @kwikteam

I have experience with Kwik format, but I'm not actively using the format.
We are using kilosort, so I think we need to build NWB files from this format anyway.
I don't know much about suite2p beyond the documentation, nor do I have example data.

[neuromusic]

Plexon Offline Spike Sorter
This is probably the most popular spike sorting software, especially for <32ch.

• modality: ephys
• output files: plx
• docs on outputs: might be best to use Neo IO here http://neo.readthedocs.io/en/0.5.0/io.html#neo.io.PlexonIO
• developers: Plexon
• note: using the Neo IO here would effectively provide a general use case for ALL ephys formats that Neo supports. see also: Add support for the NWB format NeuralEnsemble/python-neo#221

[rly]

Conversion of processed data to NWB is being handled and documented by https://neuroconv.readthedocs.io/en/main/. Processing tools are listed now in https://nwb-overview.readthedocs.io/ including some example scripts provided by the tools themselves.