Skip to content

README.md

Latest commit

 

History

History
108 lines (77 loc) · 3.44 KB

README.md

File metadata and controls

108 lines (77 loc) · 3.44 KB
Raw

scripts/

run_spe1_batch.py

Headless batch runner for the spe-1 LFP-spike cluster analysis pipeline. Replaces manual per-cell notebook execution with a single command that caches each step independently (spike fit → clustering → LFP windowing → sliding-window stats).

The script is executable, so you can call it either way:

python scripts/run_spe1_batch.py --no-plot
# or
./scripts/run_spe1_batch.py --no-plot

Quick start

# All cells, skip steps whose pickles already exist, no plots
python scripts/run_spe1_batch.py --no-plot

# Priority cells only, 4 parallel workers, force-redo LFP stats
python scripts/run_spe1_batch.py --priority --workers 4 --force-lfp --no-plot

# Specific cells
python scripts/run_spe1_batch.py --cells 21 24 42 --no-plot

# Full rerun from scratch
python scripts/run_spe1_batch.py --force-all --no-plot

Flags

Flag Effect
--cells 21 24 42 Run specific cell numbers only
--priority Run only config.PRIORITY_CELLS = [3, 21, 22, 24, 26, 28, 45]
--workers N Parallel processes (default 1 = sequential)
--force-fit Recompute spike fitting even if pickle exists
--force-cluster Recompute clustering even if pickle exists
--force-lfp Recompute LFP windowing + sliding stats
--force-all All of the above
--no-plot Skip all matplotlib rendering (much faster; stats still saved)

Path overrides

Data and pickle roots default to the paths in config.py but can be overridden without editing any files:

export SPE1_DATA_ROOT="/path/to/Neuropixel Paired Recordings/Recordings"
export SPE1_PICKLE_ROOT="/path/to/spe1_pickles"
python scripts/run_spe1_batch.py --no-plot

Pickle layout (under SPE1_PICKLE_ROOT)

spe1_pickles/
├── spike_fit_pickles/
│   └── {cnum}_spike_fit.pkl
├── cluster_pickles/
│   └── {cnum}_cluster_df.pkl
├── lfp_window_pickles/
│   └── {cnum}_lfp_windows.pkl
├── simple_lfp_pickles/
│   └── {cnum}_simple_lfp.pkl
├── multitaper_pickles/
│   └── c{cnum}/               ← specparam chunks (pre-computed, read-only)
└── lfp_spk_group_pickles/
    ├── {cnum}_sliding_stats.pkl
    └── {cnum}_per_spike_data.pkl

Each step only reruns if its pickle is missing or the matching --force-* flag is set.

run_ridge_psd_cell.py

Runs the per-cell spike-to-LFP ridge regression pipeline for a single cell. Called by the batch runner or directly for a specific cell. Loads the cluster pickle, extracts pre/post-spike LFP features, runs 5-fold CV ridge regression with permutation testing, and saves a per-cell ridge results pickle.

python scripts/run_ridge_psd_cell.py --cell 21

run_allen_ct_batch.py

Batch runner for the Allen Cell Types dataset (ground-truth cell type validation). Analogous to run_spe1_batch.py but for the allen-cell-types dataset.

save_nocluster_feats.py

Creates cluster_df.pkl for cells that have spike fit pickles but no waveform clusters (c17, c18, c43). Saves waveform features + log ISI without cluster columns so these cells can participate in non-clustering analyses (e.g. intra-spike correlations, within-vs-between waveform variability). These cells are excluded from ridge regression and all cluster-based analyses.

python scripts/save_nocluster_feats.py

benchmark_spe1_notebooks.py

Performance profiling script for the spe-1 analysis pipeline.