arista.processing.gain

Linear-fit gain per recording across the stimulus-response steps.

Reproduces Kossen 2019 Fig 27: for each recording, fit a line through all (delta_target_c, dfbf_response_median) points and use the slope as the cell’s ΔF/F-per-°C gain. CC cells produce negative slope (cold → high response), HC cells produce positive slope; the absolute value is comparable across cell types.

Gains are computed on demand from the stimulus_responses table — not persisted alongside adaptation_fits because each fit is two np.polyfit calls and the full corpus completes in well under a second. Persisting would add schema surface and a migration without saving any meaningful time at figure-render time.

Functions

`compute_gains_table`(response_df, *[, ...])	Per-recording gains across an already-fetched response frame.
`compute_recording_gain`(steps_df)	Linear fit slope of ΔF/F vs Δ target temperature.

Classes

RecordingGain(slope, intercept, r_squared, ...)

One linear-gain fit for a single recording.

class arista.processing.gain.RecordingGain(slope, intercept, r_squared, n_points)[source]

Bases: object

One linear-gain fit for a single recording.

Parameters:

slope (float)
intercept (float)
r_squared (float)
n_points (int)

intercept: float

n_points: int

r_squared: float

slope: float

arista.processing.gain.compute_gains_table(response_df, *, min_steps=3, group_keys=('recording_id', 'strain_name', 'cell_type', 'hemisphere'))[source]

Per-recording gains across an already-fetched response frame.

Parameters:

response_df (pandas.DataFrame) – Output of arista.viz.response_curves.fetch_response_data() — one row per (recording, step) with all v_recordings metadata joined in.
min_steps (int) – Drop recordings with fewer than this many valid steps (default 3 — needed for a meaningful linear fit).
group_keys (tuple[str, ...]) – Columns that uniquely identify one recording’s row group. Must include recording_id.

Returns:

DataFrame with one row per recording carrying the group keys plus slope, intercept, r_squared, n_points.

Return type:

pandas.DataFrame

arista.processing.gain.compute_recording_gain(steps_df)[source]

Linear fit slope of ΔF/F vs Δ target temperature.

Parameters:: steps_df (pandas.DataFrame) – One recording’s rows from stimulus_responses, with delta_target_c and dfbf_response_median columns.
Returns:: RecordingGain. slope and r_squared are NaN when fewer than 2 finite points are available or when every x-value is identical (no spread to fit through).
Return type:: RecordingGain