franklab_mountainsort

Installation

1. Check if miniconda or anaconda is installed https://docs.conda.io/projects/conda/en/latest/user-guide/install/linux.html

2. Clone git repository

git clone https://github.com/LorenFrankLab/franklab_mountainsort.git

3. Go into the franklab_mountainsort folder

cd franklab_mountainsort

4. Create conda environment

conda env create -f environment.yml

5. activate conda environment

conda activate franklab_mountainsort

6. Install package code

python setup.py develop

7. Check if everything installed correctly. Open jupyter notebook, jupyter lab, or python console. Try to import franklab_mountainsort.

import franklab_mountainsort

Usage

1. Define input data directory, output data, temp directory
2. Create a pandas dataframe with .mda file information (using function get_mda_files_dataframe)
3. Run spike sorting on all or a subset of the DataFrame (using function spike_sort_all)
   • This will concatenate all epochs in a day
   • Bandpass filter and whiten the data
   • Run the isosplit clustering algorithm
   • Merge clusters that look like they are from one cell that is bursting (optional).
   • Add curation metrics and tags (statistics about each cluster and whether it is multiunit or not.).
   • Extract clip waveforms (optional)
   • Extract marks (optional)
4. Open qt-mountainview (make sure you've activated the conda environment. If not type conda activate franklab_mountainsort into the terminal.)

qt-mountainview --pre=pre.mda.prv --firings=firings_raw.mda --samplerate=30000 --cluster_metrics=metrics_raw.json

Usage Note

• Make sure the version of trodes you used to extract the data is the same that you recorded with.

What does spike_sort_all do?

For each set of channels on an electrode:

1. Concatenate timeseries over epochs
2. Bandpass filter
   • between 300 and 6000 Hz by default
   • relevant arguments: freq_min, freq_max, sampling_rate
3. Mask out high voltage artifacts
   • remove chunks of data with amplitudes outside of the expected range by replacing these data by zeros
   • probably should expose these parameters
4. Whiten
   • decorrelate signals

If drift_track:

5. Identify clusters on each epoch individually
   • using the isosplit clustering algorithm
   • relevant arguments: adjacency_radius, detect_threshold, detect_interval, detect_sign, num_workers
6. Anneal segments
   • stich epochs back together while matching clusters from the different epochs
   • compares identified clusters across all epochs

7. Compute cluster metrics
   • firing rate
   • isolation: how well separated (in feature space) the cluster is from other nearby clusters
   • noise_overlap: estimates the fraction of “noise events” in a cluster, i.e., above-threshold events not associated with true firings of this or any of the other clustered units.
   • peak_snr: peak signal to noise ratio

If burst_merge:

8. Merge burst parents
   • automatically merge clusters that are demeed to have come from the same bursting cell

9. Add curation tags
   • Add tags to the metrics file to reflect which clusters should be rejected based on curation criteria.
   • rejected, accepted, `mua```
   • relevant arguments: firing_rate_thresh, isolation_thresh, noise_overlap_thresh, peak_snr_thresh

If extract_marks:

10. Extract marks
    • extract the peak amplitude at the time of the spike

If extract_clips:

11. Extract clips
    • extract the waveforms around the time of the spike
    • relevant arguments: clip_time, sampling_rate

Inputs to MountainSort

• *.mda: time series for each recording.
• params.json: contains information about the parameters to use for the sort. These can be overwritten by specifying them in the call to run the sort (in the batch script)
• geom.csv (optional): contains information about the location of contacts for that ntrode; used in concert with adjacency_radius to determine the neighborhoods to sort on. In the case of tetrodes, this is not necessary because all the contacts of a tetrode should be sorted together as a single neighborhood. This can be specified by not providing a geom.csv, setting adjacency_radius to -1, or both.

Note: .prv vs..mda files

A .prv file is a text file with pointers to the binary file. A .mda file is a binary file containing the actual data.

They can be used interchangeably when using qt-mountainview.

Outputs from MountainSort

For each electrode:

• raw.mda.prv: The concatenated time series for one day of recording. It is located in the ../<animal>/mountainlab_output/<date>/<date>_<animal>.mountain folder
• clips.mda: The waveforms around the spike times. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• filt.mda.prv: The time series after it has been bandpass filtered. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• firings_burst_merged.mda: The firings file after burst merge processing. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• firings_raw.mda: This contains the actual spike timing info that you care most about [electrode; time;label x #events….] for ALL detected events, regardless of cluster quality. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• metrics_merged.json: This contains the metrics for the curated clusters, such as isolation scores, noise overlap, SNR, and more. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• metrics_merged_tagged.json: Same as metrics merged but with tags. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• metrics_raw.json: Metrics for all the original clusters. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.
• pre.mda.prv: The time series after it has been bandpass filtered and whitened. Located in the ../<animal>/mountainlab_output/<date>/nt<number>/ folder.