plx-to-kilosort

A bit of Matlab code to convert Plexon .plx files into formats we can pass to Kilosort and/or Phy.

You can run this code locally and/or in containers.

Containers

Two container images are produced from this repo and shared on Docker Hub:

• ninjaben/plx-to-kilosort converts a Plexon .plx file to a raw .bin file, plus Kilosort "ops" and "chanMap" files, in both '.mat' and '.json' format for convenience. The results of this step can be passed on to Kilosort for sorting.
• ninjaben/plx-to-phy can take the .plx, .bin, and "ops" produced above and convert directly to "Phy" format. This allows review and curation of spike sorting results that were performed manually with Plexon and saved in the .plx file. This is intended to allow comparison of manual and automatic spike sorting using the same Phy tool.

Here are some example Docker commands for running these in containers.

Conversion of .plx file to .bin and "ops"

This step uses Matlab inside the container.

LICENSE_MAC_ADDRESS=$(cat /sys/class/net/en*/address)
LICENSE_FILE="$(pwd)/license.lic"
sudo docker run --rm \
  --mac-address "$LICENSE_MAC_ADDRESS" \
  --volume $LICENSE_FILE:/licenses/license.lic \
  --env MLM_LICENSE_FILE=/licenses/license.lic \
  --volume "/path/to/my/plexon/data:/plexon/data" \
  ninjaben/plx-to-kilosort:v0.0.13 \
  -batch "[chanMapFile, binFile, opsFile] = plxToKilosort('/plexon/data/recording.plx', '/plexon/data/out', 'tRange', [0, 30], 'ops', {'fproc', '/plexon/data/scratch/temp_wh2.dat'})"

Conversion of .plx,.bin, and "ops" to Phy

This step uses spikeinterface and Jupyter.

batch mode

sudo docker run --rm \
  --volume "/path/to/my/plexon/data:/plexon/data" \
  --env PLX_FILE="/plexon/data/recording.plx" \
  --env BIN_FILE="/plexon/data/out/recording.plx.bin" \
  --env OPS_FILE="/plexon/data/out/recording-ops.json" \
  --env OUT_DIR="/plexon/data/out/" \
  ninjaben/plx-to-phy:v0.0.13  \
  jupyter nbconvert --execute /home/jupyter/notebooks/plx_to_phy.ipynb --to html"

interactive mode

sudo docker run --rm \
  --volume "/path/to/my/plexon/data:/plexon/data" \
  --network=host \
  ninjaben/plx-to-phy:v0.0.13  \
  "jupyter notebook"

Then visit localhost:8888.

Without Containers

Matlab setup

The code in the plx-to-kilosort/matlab folder can be run locally in Matlab. To run it locally you need to insteall a dependency -- the Plexon OmniPlex and MAP Offline SDK Bundle. This is available from the Plexon Software Downloads page (accessed December 2022).

Once you have the OmniPlex and MAP Offline SDK Bundle:

• Unzip it.
• Find Matlab Offline Files SDK.zip within.
• Unzip that, too.
• Add OmniPlex and MAP Offline SDK Bundle with subfolders to your Matlab path.
• In Matlab, execute build_and_verify_mexPlex to compile the mexPlex function.

Once that works, add plx-to-kilosort/matlab to your Matlab path and you should be ready to proceed.

Summarize .plx File in Matlab

We can summarize the contents of a Plexon .plx file with summarizePlxFile.m. This will return some header info and counts for data channels including spikes and timestamps, continuous AD "slow" channels, and digital events.

For example, to summarize 30 seconds of data starting at 100s:

>> [header, counts] = summarizePlxFile(plxFile, 100, 30);

Timestamps and waveforms:
  113101 timestamps for spike channel 2, unit 0
  58887 timestamps for spike channel 2, unit 1
  449384 timestamps for spike channel 4, unit 0
  113101 waveforms for spike channel 2, unit 0
  58887 waveforms for spike channel 2, unit 1
  449384 waveforms for spike channel 4, unit 0
Digital events:
  1 events for event channel 1 -- Event001 
  40351 events for event channel 257 -- Strobed  
  1 events for event channel 258 -- Start    
  1 events for event channel 259 -- Stop     
AD channels:
  3423016 samples for continuous / slow channel 17 -- AD18 
  3423016 samples for continuous / slow channel 47 -- AD48 
  3423016 samples for continuous / slow channel 48 -- Pupil
  3423016 samples for continuous / slow channel 49 -- X    
  3423016 samples for continuous / slow channel 50 -- Y    
  3423016 samples for continuous / slow channel 51 -- AD52 

We get Plexon header-level data:

>> header

header = 

  struct with fields:

                  file: 'myData.plx'
               version: 107
             frequency: 40000
               comment: ''
            trodalness: 1
         pointsPerWave: 50
    pointsPreThreshold: 8
            spikePeakV: 3000
           spikeAdBits: 12
             slowPeakV: 5000
            slowAdBits: 12
              duration: 3.423013875000000e+03
              dateTime: ' 8/ 5/2022 12: 2:30'

We get counts of spike timestamps and waveforms, ad samples, and digital events:

>> counts

counts = 

  struct with fields:

      tscounts: [27×17 double]
      wfcounts: [27×17 double]
      evcounts: [1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 40351 1 1 0 0 0 0 0 0 0 0 0]
    contcounts: [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3423016 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3423016 3423016 3423016 3423016 3423016 0 0 0 0 0 0 0 0 0 0 0 0]

We also get four plots, each showing a different channel type over the requested startTime and duration. Some examples are below.

Spike channel Waveforms over time

Spike channel Waveforms aligned in the trigger window

Continuous AD AKA "slow" channels

Digital event channels

Spikeinterface and Jupyter Setup

The code in the plx-to-phy/matlab folder can be run locally in a Jupyter notebook. A straightforward way to do all the setup is:

• Install conda on your machine.
• Go to the command line, cd to the plx-to-phy folder of this repo, and run the following:

conda env create -f environment.yml
conda activate si_env

From there you should be able to run the notebook in batch mode or interactively.

batch mode

PLX_FILE="/path/to/my/plexon/data/recording.plx" \
BIN_FILE="/path/to/my/plexon/data/out/recording.plx.bin" \
OPS_FILE="/path/to/my/plexon/data/recording-ops.json" \
OUT_DIR="/path/to/my/plexon/data/out/" \
jupyter nbconvert --execute plx_to_phy.ipynb --to html

interactive mode

jupyter notebook

Then visit localhost:8888.