Hi everyone. Does anyone know the time vector concerning the average waveforms that out made by Kilosort2 for the Brain Observatory datasets for the visual and resting state animals (~58 animals in total)? Also is there cell type information concerning all those units? Thanks, Atanas
The time vector for the waveforms is shown in this example notebook (cells 30 and 31). The total duration of each waveform is around 2.7 ms.
We haven’t included explicit cell type information in this dataset. To classify cells based on the results of the optotagging stimulus run at the end of each session, you can follow the steps in this notebook. This can be used to identify Pvalb, Sst, or Vip interneurons in different sessions.
1 Like
Hello thank you for the quick reply! I was wondering when the spike sorter typically demarcates zero for when the unit fired (usually trough min)?
That’s very helpful for the cell labels! Thanks again.
Sorry, I had another question. Is any “noise cluster” information available like spike times of rejected templates and/or templates marked as MUAs.
I believe Kilosort aligns timestamps to the trough minimum, which occurs 20 samples after the start of the spike waveform.
The spike trains for all “noise” units are available in the NWB files for this dataset, but they are hidden by default if you’re using the AllenSDK. To access them, you’ll have to set filter_by_validity = False when creating the EcephysSession object.
We don’t classify units as MUA, since there’s not an obvious quality threshold that separates multi-unit from single-unit activity. For more info on the quality metrics we use, you can check out this tutorial.
Thank you! Are the “noise” units accessible via the MATLAB Brain-Observatory-Toolbox? Asking because I’ve been having good success with MATLAB toolbox with the data downloads. EDIT: grammar
I’m not sure. If you’re having trouble finding this info via that toolbox, you can always open up the NWB files directly using the MatNWB library.
1 Like
No worries and thank for all the feedback! I think the Allen Institute has used some custom modifications to the NWB files and the MatNWB isn’t working with that yet. I think my workaround will be to download the datasets in parallel with python and just pull the data I need from there using the AllenSDK tools and convert the noisy units to “.mat” files within python. I noticed the optotagging data isn’t yet available via the MATLAB toolbox so I’ll need to pull that with the AllenSDK tools and convert to “.mat” there as well.