Fast notebook-based access for three P3 dandisets (Neuropixels, mesoscope, SLAP2)

[aaronkanzer]

Hi all,

@jeromelecoq suggested I share this here. I work on @Scigantic (https://scigantic.com), a site that lets you open public datasets in a hosted Jupyter notebook without downloading anything. I onboarded the three P3 dandisets and wrote a short example notebook for each:

• Neuropixels (001637): https://scigantic.com/archive/383ee6c3-4bb8-5b26-8cd3-eabd7647673f/sample
• Mesoscope (001768): https://scigantic.com/archive/0eefd34a-3522-52b5-bd03-12d5c331bb94/sample
• SLAP2 (001424): https://scigantic.com/archive/97f326e3-a79b-5202-9c36-01105c4dc13e/sample

Each notebook starts with its own workspace, then loads one session from each of the other two. The point is to show that the block design is the same across all three (same block types, same trial counts) and to line up the response to the same grating as spikes, calcium, and glutamate side by side. The links above show the pre-run notebook. Hitting "Launch" opens a live copy with all three dandisets mounted so you can edit and rerun it.

As an aside, a few things I ran into while writing these notebook examples that was interesting:

• The stimulus tables have the same content on all three platforms but slightly different names. Neuropixels and mesoscope use _presentations, SLAP2 uses the block type directly. Keying on BlockType works everywhere.
• Orientation is in radians.
• The SLAP2 dandiset has three file layouts (production, pilot, and a few older sessions), so a loader that works on one file may not work on the next.

If I got anything wrong about the data, please tell me and I'll fix the notebooks. Happy to add anything else that would be useful to the group.

Aaron