QC and Analysis Pipeline

The IPFX package is utilized in the Allen Institute’s electrophysiology processing pipeline that is being used to generate publicly available data.

Given the nwb2 file with electrphysiological recording for a single cell the pipeline will perform the following steps (run with the corresponding executables):

1. Compute QC features (ipfx.bin.run_sweep_extraction)

   Extract sweeps and their metadata from the nwb file Computed QC features for the cell as a whole and also for each individual sweeps

2. Perform QC checks (ipfx.bin.run_qc)

   Check QC criteria on QC features at the level of the entire cell and for individual sweeps Tag cell, sweeps failing QC criteria

3. Compute intrinsic features (ipfx.bin.run_feature_extraction)

   Compute features of spikes, spike trains, as well as stimulus specific features

4. Attach metadata (ipfx.attach_metadata)

   Add ancillary information about an experiment to the output nwb2 file

Each executable defines a schema for the input parameters specified in the <input.json> and can be invoked as:

python -m <executable> --input_json <path/to/input.json> --output_json <path/to/output.json>

where <executable> stands for the executables listed in the above pipeline steps.

Running the pipeline requires two additional pieces of information:

1. Stimulus ontology that maps names of sweeps in the input nwb2 file to the stimulus types known to ipfx
2. QC criteria that specify the acceptable values for the computed QC features

If not explicitly provided, the pipeline will invoke the default values from the ipfx.defaults folder