In vivo Ephys: Spike Sorting Protocol

For more information on setting up and using Spyking Circus, see: https://spyking-circus.readthedocs.io/en/latest/code/index.html .

Make a copy of the continuous.dat data file from your recording into a new “sorted_data” folder.

Add a .params file with the same name as the .dat file to the folder (i.e., continuous.params). The continuous.params used for these experiments:

[data]
file_format    = raw_binary
data_offset    = 0                    
mapping        = path\open_ephys_16prb.prb     # Mapping of the electrode 
							                   #see prb file below
suffix         =           # Suffix to add to generated files
data_dtype     = int16     # Type of the data
dtype_offset   = auto      # Padding for data (if auto: int16 is 0)
skip_artefact  = False     # Skip part of signals with large fluctuations
sampling_rate  = 30000     # Sampling rate of the data [Hz]
stationary     = True      # Should be False for long recordings
radius         = auto      # Radius [in um] (if auto, read from the prb file)
alignment      = True      # Realign the waveforms by oversampling
global_tmp     = True          
multi-files    = False     # If several files mydata_0,1,..,n.dat should be processed together

[detection]
spike_thresh   = 4
N_t            = 2 or 3    # Width of the templates [in ms]. We used both 2 and 3 in different sortings to extract all the templates included.
peaks          = positive  # can be positive, negative, or both

[filtering]
cut_off        = 250       # Cut off freq for the butterworth filter [Hz]
filter         = True      # If True, then a low-pass filtering is performed 

[whitening]
chunk_size     = 60        # Size of the data chunks [in s]
safety_time    = 1         # Temporal zone around which templates are 
isolated [in ms]
temporal       = False     # Perform temporal whitening 
spatial        = True      # Perform spatial whitening
max_elts       = 10000     # Max number of events per electrode
nb_elts        = 0.8       # Fraction of max_elts that should be obtained per electrode [0-1]
output_dim     = 5         # Can be in percent of variance explain, or num of dimensions for PCA on waveforms

[clustering]
extraction     = median-raw # Can be either median-raw (default), median-pca, mean-pca, mean-raw, or quadratic
safety_space   = True      # If True, we exclude spikes in the vicinity of a selected spikes
safety_time    = auto      # Temporal zone around which templates are isolated [in ms]
max_elts       = 15000     # Max number of events per electrode (should be compatible with nb_elts)
nb_elts        = 0.8       # Fraction of max_elts that should be obtained per electrode [0-1]
nclus_min      = 0.0075    # Min number of elements in a cluster (given in percentage)
max_clusters   = 20        # Maximal number of clusters for every electrodes
nb_repeats     = 3         # Number of passes used for the clustering
smart_search   = 0         # Parameter for the smart search [0-1]. The higher, the more strict
sim_same_elec  = 0.75      # Distance within clusters under which they are re-merged
cc_merge       = 1         # If CC between two templates is higher, they are merged
noise_thr      = 0.8       # Minimal amplitudes are such than 
amp*min(templates) < noise_thr*threshold 
make_plots     = png       # Generate sanity plots of the clustering 
remove_mixture = True      # At the end of the clustering, we remove mixtures 
of templates

[fitting]
chunk          = 1         # Size of chunks used during fitting [in second]
gpu_only       = False     # Use GPU for computation of b's AND fitting
amp_limits     = (0.3, 5)  # Amplitudes for the templates during spike 
detection
amp_auto       = True      # True if amplitudes are adjusted automatically 
for every templates
refractory     = 0         # Refractory period, in ms [0 is None]
max_chunk      = inf       # Fit only up to max_chunk   

[merging]
cc_overlap     = 0.5       # Only templates with CC higher than cc_overlap 
may be merged
cc_bin         = 2         # Bin size for computing CC [in ms]

[extracting]
safety_time    = 3         # Temporal zone around which spikes are isolated 
[in ms]
max_elts       = 1000      # Max number of collected events per templates
output_dim     = 5         # Percentage of variance explained while 
performing PCA
cc_merge       = 1         # If CC between two templates is higher, they are 
merged
noise_thr      = 0.8       # Minimal amplitudes are such than
amp*min(templates) < noise_thr*threshold

[noedits]
filter_done    = False     # Automatically edited – make sure this is False 
before running the sorter

The .params file needs to have correct path to a .prb file (the mapping of the electrode you are using). We recorded with bundles of 16 individual electrodes, and used the following .prb file:

total_nb_channels = 16
radius            = 80
channel_groups = {
    1: {
        'channels': list(range(16)),
        'graph' : [],
        'geometry': {
		 0: (0, 300),
		 1: (100, 300),
		 2: (200, 300),
		 3: (300, 300),
		 4: (0, 200),
		 5: (100, 200),
		 6: (200, 200),
		 7: (300, 200),
		 8: (0, 100),
		 9: (100, 100),
		 10: (200, 100),
		 11: (300, 100),
		 12: (0, 0),
		 13: (100, 0),
		 14: (200, 0),
		 15: (300, 0),
        }
    }
}

Launch the Spyking Circus GUI from the anaconda terminal.

Run the sorting in the Main Algorithm tab.

Browse → select continuous.dat file.

Increase CPUs as desired, then click “Run”.

When the next window pops up, click “Suggest pairs”, “Select pairs”, then “Merge and finalize”.

To view results: View Results tab → Browse → select the analyzed continuous.dat file.

Options:

1. Matlab GUI
2. Extension = “merged”

Click “Run”.

See https://spyking-circus.readthedocs.io/en/latest/GUI/sorting.html for detailed documentation on what each panel of the GUI does.

Sorting:

Set bin size to 1, click “Big ISI”, and update the scale so you can see whole waveforms (zooming out usually works).

Go through all templates.

First pass: Kill a cell if it has above a 1% incidence of refractory period violations.

Second pass: merge templates that you think are the same cell, using the two template view.

• Criteria used: similar waveforms, cross correlation dropping to zero at zero, coefficient of similarity close to 1.
• Sometimes you have to merge more than two templates for one cell.
• Merge template with fewer spikes into the one with more spikes.

Third pass: kill units that don’t meet requirements (too sparse, not spiking consistently across the whole session, auto correlation doesn’t go to zero, RPV has gone above 1%).

Add suffix (ex, ‘v1’ for first attempt at sorting) and save.