python-codebase-for-extracellular-recording-simulation

Codebase structure

• algo (folder to save different sorting alogrithm source code)
  • algoname1
    • ...
  • algoname2
  • ...
• cell_models (folder to save raw cell model groups from Allen)
  • cellmodelsgroup1
  • cellmodelsgroup2
  • ...
• simulate (code for generating simulated datasets)
• compare (code for comparison)
• studies (folder to save several studies)
  • studyname1
    • cell_sim_model (folder to save simulated cell models in this study)
    • templates (can generate several templates from one group of cell_model, folder to save these templates)
      • templatesname1
        • templates.h5 (from MEArec.save_template_generator(), has params)
        • probename.yaml (a copy of probe file used in generating templates by MEArec)
        • template_params.pkl (params to generate this templates.h5)
      • templatesname2
      • ...
    • recordings (can generate several recordings from one set of templates, folder to save these recordings)
      • recordingname1
        • recordings.h5 (from MEArec.save_recording_generator(), has params)
        • recordings.bin (extract only raw data as int16 bin file, as some sorting algos require)
        • recording_params.pkl (params to generate this recordings.h5)
        • info.pkl (general infomation of this recording, like path to cell_models, cell_sim_models, templates)
        • probe (folder to save probe files with different formats, as some sorting algos require; should have the same information to the probe file in template folder)
          • probeforalgo1.prb
          • probeforalgo2.mat
          • ...
        • sorting_results (folder to save all algos result files)
          • algoname1
          • algoname2
          • ...
        • comparing_results
          • fig (folder to save figure plots)
            • ...
          • ...
      • recordingname2
      • ...
  • studyname2
    • ...

Comparing results

In general, we can have 2 categories of comparing results: spike time comparing and template waveform comparing.

Spike time comparing

• time_agreement_matrix()
• jaccard_score_matrix()
• ccg_matrix() (from pykilosort)

Template waveform comparing

• cosine_similarity_matrix()
• simscore_matrix() (svd, similar to pykilosort)

About my study mouse_VISp_L5_128ch

This study was created for a systematic comparison between some different algorithms. The neural models are from The Allen Cell Type Database.

Algorithms:

• kilosort2.5 (KS25)
• kilosort3 (KS3)
• herdingspikes (HS2)
• spykingcircus (SC)
• spykingcircus2 (SC2)
• tridesclous (TC)
• tridesclous2 (TC2)
• mountainsort5 (MS5)

Generated recordings (simulated datasets):

Condition	recording_toy	recording_onlydrift_slow	recording_onlydrift_fast	recording_drift
Jitter	$\surd$	$\surd$	$\surd$	$\surd$
Shape modulation	$\times$	$\times$	$\times$	$\surd$
Bursting	$\surd$	$\surd$	$\surd$	$\surd$
Noise	distance-correlated	distance-correlated	distance-correlated	distance-correlated
Drifting	$\times$	slow	fast	slow & fast