Functions that create intermediate analysis files. This is meant to work on a file hierarchy compatible with marmolab-pipeline.
It integrates:
- kilosort cluster outputs and spike times
- digital event timing from NEV for stimulus synch
- information from Stim files (.mat)
There are slurm files to run the batching on the cluster, used in the following order:
runSpikeExtractNeurostim.sbatch extracts spikes for each channel in an marmodata file and saves them in spike train format by calling RUNAnalysis_mdbExtract.m. The congfiguration for which files this processes is provided in configureBatch_mdb.m. This isn't necessary for the data recorded on Blackrock systems.
runProcessSpiking.sbatch calls RUNAnalysis_brain.m to merge files and channels into aggregate spike trains, tuning curves, sdfs, and exclusion files. It produces the following intermediate data files that are useful for further analysis.
Use tests/configureRunTests.m to generate good summaries of the exports so that we can assess the quality of the data/detect bugs or encoding problems in the data files.
chanOrder (1x2 cell, where each cell is an array/area)
- has a list of all the channels, for the NN arrays in depth order, for Utah arrays, just 1-96 (see #1 ). However, this is kind of pointless for data post-KS. With KS this data is in
clustInfo{i}.clusterInfo
clustInfo (1x2 cell, where each cell is an array/area)
- has the kilosort related information for each unit identified in each area.
onsetInds (1x4 cell, where each cell is a 1x2 recording file x array)
This has the stimulus identity and timing.
param (structure)
This has information about the trial types and time course, and which integration windows should be used for 1) spike counts or 2) SDFs.
StimFile (1x4 cell, with each cell a recording file)
This has the information used to create the stimulus, bundled from the separate stimulus file saved during the experiment.
sTrain (1x4 cell, with each cell a recording file, where each cell is a 1x2 array index)
This has spike trains for each neuron or electrode during the recording. Millisecond resolution, 1/0 encoding.
chanOrder (1x2 cell, where each cell is an array/area), param (structure)
As above. Re-encoded here, because this file is more lightweight.
tcs (structure)
tcs.[Move | Blank].[Dots | SineWave | Square | PSsquare]{array, direction}(2x12 cell array). Each cell has a nUnitsxnTrials spike count, integrated during the window indicated in param.
Zscs (structure)
Mirror of tcs, but has z-scored rates not raw rates. Z-scoring is done within a stimulus type so stimulus selectivity is destroyed.
tcs_byFile (structure)
tcs.[Move | Blank].[Dots | SineWave | Square | PSsquare]{1x4}{array, direction}
Tuning curves, but without the trials combined over different recording files.
isVisual (1x2 cell, where each cell is an array/area)
Has 4xnUnits array where visual responsiveness was tested separately for each stimulus type. 1 = has significant visual response.
anyVisual (1x2 cell, where each cell is an array/area)
As above, but 1= has significant visual response for any of the stimulus types.
DSI (1x2 cell, where each cell is an array/area)
Direction selectivity for any unit with a significant visual response. Difference/sum of the peak and anti-peak direction.
param (as elsewhere)
all_sdfs
all_sdfs.[Dots | SineWave | Square | PSsquare]{array, direction}
Each cell is a nChannelsxnTrialsxnTimepoints 3D matrix.