Skip to content

Latest commit

 

History

History
142 lines (99 loc) · 6.04 KB

README.md

File metadata and controls

142 lines (99 loc) · 6.04 KB
Raw

BrainCogsEphysSorters

Compilation of electrophysiology sorters and preprocessing tools supported in the U19 Ephys Pipeline

User documentation

What to run to test the code:

Example of a slurm command to preprocess and process an ephys session (Normally all parameters will be generated from the automation pipeline):

sbatch --export=recording_process_id=28,raw_data_directory='ms81/ms81_M004/20210507/towersTask_g0/towersTask_g0_imec0',processed_data_directory='ms81/ms81_M004/20210507/towersTask_g0/towersTask_g0_imec0/recording_process_id_28',repository_dir='/scratch/gpfs/BRAINCOGS/electrophysiology_processing/BrainCogsEphysSorters',process_script_path='main_script.py' slurm_real.slurm

To test code for a new ephys session:

  1. Go to directory /scratch/gpfs/BRAINCOGS.
  2. Check that the ephys session you want to process is located on /Data/Raw/electrophysiology/(netID)/subject_fullname/date/.. (ask main developer if session is not there).
  3. Copy parameters files (preprocess_paramset_x.json & process_paramset_x.json)located in /ParameterFiles (x is a number that reference the ephys session).
  4. Rename copied parameter files with a new number (preprocess_paramset_y.json & process_paramset_y.json) (y could be any number)
  5. Run sbatch command (from the top of this section) with following modifications:
  • recording_process_id =y (Change the number to match the parameter file name modification
  • raw_data_directory = Relative directory to the ephys session probe (netID)/subject_fullname/date/.../imec(z)
  • processed_data_directory = Same as raw_data_directory but include recording_process_id_y (where y is the number from the parameter file)

Logs will be written:

  • for Kilosort: /Output_log/kilosort_out.log
  • for CatGT: /BrainCogsEphysSortets/CatGT.log (appended every time)

Preprocess parameter file

Preprocess parameter file is a json file to configure preprocessing steps and select the desired sorter, main parameters are:

  • clustering_method: (str) Which sorter will be used (Kilosort2, Kilosort, SpikeInterface).
  • cat_gt: (dict) All parameters corresponding to catGT preprocessing.
    • use_cat_gt: (bool) 0 if cat_gt will not be used, 1 otherwise.
    • "cat_gt_params": (dict) Cat_GT params that are included in command line when it is called
  • process_cluster: (str) NA, Choose which available cluster will be used for processing (tiger, spock, etc)

Example: Next preprocess param file will run Kilosort2 and use catGT:

{
"process_cluster": "tiger",
"clustering_method": "Kilosort2",
 "cat_gt": {
    "use_cat_gt": 1,
    "cat_gt_params": {
      "dir": "",
      "run": "towersTask",
      "g": 0,
      "t": 0,
      "apfilter": ["biquad",2,300,0],
      "gfix": [0.40,0.10,0.02],
      "extras": ["prb_3A", "prb_fld", "t_miss_ok", "ap", "gblcar", "out_prb_fld"],
      "dest": ""
   }
 }
}

The catGT command from these params is:

./runit.sh '-dir=raw_data_directory -run=towersTask -g=0 -t=0 -prb_fld -prb=0 -t_miss_ok -ap -apfilter=biquad,2,300,0 -gblcar -gfix=0.40,0.10,0.02 -dest=processed_data_directory -out_prb_fld

Process parameter file

Process parameter file is a json file to configure sorter. To configure this file refer to each sorter documentation:

Tiger cluster directory organization

As BRAINCOGS we have a common directory in the Tiger Cluster for processing data:

/scratch/gpfs/BRAINCOGS

These are the main directories that can be found on it:

  • /Data → Replicate /braininit/Data directory path (raw sessions, processed sessions, etc)

  • /electrophysiology_processing → All repositories for processing ephys

    • /BrainCogsEphysSorters Common library to process ephys in all BrainCogs (this Repository)

  • /ParameterFiles → Preprocess and process params files for processing sessions (all modalities)

  • /Output_log → Output data from processing sessions (all modalities)

Repository directory organization

These are the main directories of the repository

  • /preprocess_libs → All libraries we are going to use for preprocessing (CatGt, etc)

  • /sorters → All repositories from sorter algorithm libraries (Kilosort, SpikeInterface, etc.)

  • /u19_sorting → Call to all preprocess and process (sorters) codes.

    • preprocess_wrappers.py Wrappers to call all preprocess libraries with corresponding params
    • sorters_wrappers.py Wrappers to call all sorters libraries with corresponding params

  • main_script.py → Script that is executed on runtime

Main developer documentation

Instructions to clone and setup the repository

Since this is a repository with submodules it is needed to add --recurse-submodules when cloning it. So log into tiger and execute:

git clone --recurse-submodules [email protected]:BrainCOGS/BrainCogsEphysSorters.git

And for branches, add the flag

git clone -b tmp --single-branch --recurse-submodules https://github.com/BrainCOGS/BrainCogsEphysSorters

Set up CatGT

For preprocessing, we currently support CatGT:

cd ./CatGT-linux/
chmod +x ./install.sh
./install.sh

Revised sorter:

  1. User config files (channel maps and sorting config files) should be deposited in user_config_files.

  2. The two example channel maps chanMap_npx1_staggered.mat and chanMap_npx2_hStripe_bottom2.mat were written for staggered Neuropixel 1.0 probes, and the bottom horizontal strip of a 4-shank Neuropixel 2.0 probe.

  3. The example config file config_manuel.m is adjusted and mildly optimized from the file \eMouse_drift\config_eMouse_drift_KS2.m, which is part of the kilosort repository.

  4. A test script that uses kilosortbatch, a wrapper around kilosort, is in the folder /sandbox/. It plays well with npy-matlab version b7b0a4e and kilosort version 1a1fd3a.

  5. Once kilosort has run, waveforms should be inspected with phy, e.g. phy template-gui D:\NPX_DATA\manuel\tmp\TowersTask_g0_imec2\params.py

  6. TODO: ks2_run is not needed anymore. Some general cleanup.