modules

• dlab.pprox: functions for working with pprox objects, a data format for storing multi-trial point process data (e.g. spike times evoked by stimulus presentation).

• dlab.neurobank: some convenient wrappers for interacting with a neurobank repository and archives.

• dlab.signal: signal processing functions

console scripts

• group-kilo-spikes: sort spike times output from kilosort and phy2 into pprox files (stimtrial format), optionally recording pulses on auxiliary channels such as optogenetic stimulation (--aux). For recordings without a sync track, --audio-sync CHANNEL finds the stimulus onsets on a copy of the audio by cross-correlation (python -m dlab.audio_sync does this on its own, for checking a recording). If only part of the recording was sorted, --sort-start and --sort-stop give the part. Where kilosort’s output for it starts is found by checking it against the recording’s neural channels, to the sample, and the output is timed from the recording’s start. Stimuli presented under other names than they were deposited under are mapped with --alt-stim-names. The outputs record the sort, the sorter’s version (read from kilosort 4’s log, or given with --sorter), and how many of each unit’s spikes were dropped and why.

• audit-kilo-spikes: check the group-kilo-spikes output for one recording against its waveform files, the stimulus messages in the ARF file, and each other, and write a JSON report. Changes nothing.

• find-kilo-units: find the units of recordings in the registry and write a control file for running audit-kilo-spikes in batches.

• collect-kilo-audit: summarize the audit-kilo-spikes reports.

• regenerate-pprox: rebuild missing pprox files from their waveform files.

• get-songs: extract segments from arf files, rescale, resample, save into wave files, and optionally deposit back into neurobank. Usually the first step in generating a stimulus set, keeps a nice provenance trail.

See docs/audit.md for how to run the audit and what its findings mean.

other stuff