Data from: Intrinsic properties link a network model to zebra finch song

Published Jul 24, 2026 on Dryad. https://doi.org/10.5061/dryad.wdbrv162x

Data files

Jul 24, 2026 version files 9.81 GB

Abstract

Neuronal intrinsic excitability is a mechanism implicated in learning and memory that is distinct from synaptic plasticity. Prior work in songbirds established that intrinsic properties (IPs) of premotor basal-ganglia-projecting neurons (HVC_X) relate to learned song. Here we find that temporal song structure is related to specific HVC_X IPs: HVC_X from birds who sang longer songs including longer invariant vocalizations (harmonic stacks) had IPs that reflected increased post-inhibitory rebound. This suggests a rebound excitation mechanism underlying the ability of HVC_X neurons to integrate over long periods of time and represent sequence information. To explore this, we constructed a network model of realistic neurons showing how in-vivo HVC bursting properties link rebound excitation to network structure and behavior. These results demonstrate an explicit link between neuronal IPs and learned behavior. We propose that sequential behaviors exhibiting temporal regularity require IPs to be included in realistic network-level descriptions.

Publication Information

Medina, N. D., & Margoliash, D. (2026). Intrinsic properties link a network model to zebra finch song. eLife. https://doi.org/10.7554/eLife.99611

Author Information

Contact: Nelson D. Medina
Email: NelsMedina010@gmail.com

Dataset Overview

This dataset contains electrophysiological data and summary statistics for HVC projection neurons (HVC-RA and HVC-X) recorded in brain slices from male zebra finches, together with a MATLAB model file and a Python companion notebook.

File List & Descriptions

1. Python Data Files (Pickle Format)

all_birds_HVC_RA.pickle: Extracted electrophysiological features and/or trace data for HVC-RA neurons.
all_birds_HVC_X.pickle: Extracted electrophysiological features and/or trace data for HVC-X neurons.

Note: These are Python-specific binary files containing nested dictionaries, organized as bird → cell → protocol (e.g. 'test') → per-sweep feature arrays, alongside the injected-current commands (neg_command, pos_command) and sweep times.

2. Cardinal_2.0.zip (zipped "Cardinal 2.0" folder)

Summary statistics in Excel (.xlsx) format.

Organization: one .xlsx file per subject (bird).
Structure: each tab within a file corresponds to a single recorded cell.
Contents: per-cell summary intrinsic-property metrics used for the figures in the paper. All variables are defined in the "Variable definitions (Cardinal_2.0 spreadsheets)" section below.

3. example_analysis.ipynb (companion Python notebook)

A minimal, self-contained Jupyter notebook demonstrating how to work with the dataset. It:

loads the two .pickle files;
documents their nested-dictionary structure (bird → cell → protocol → per-sweep feature arrays);
reproduces the paper's core feature-extraction routine (first_n_traces: select the sweeps recorded at a target injected-current command, drop non-finite values, and take the median of the first n sweeps per cell); and
shows a representative reproducible analysis — extracting sag ratio for HVC-X vs HVC-RA cells and comparing their distributions with a Mann–Whitney U test.

It uses only standard scientific-Python packages (numpy, pandas, scipy, matplotlib, seaborn). It is an illustrative guide to the data format, not a one-click regeneration of every figure. To run it, place the notebook in the same folder as the .pickle files (and HVC_model_rebound_only.mat), or edit the file paths in the notebook.

4. requirements.txt

Plain-text list of the Python package dependencies needed to run example_analysis.ipynb (numpy, pandas, scipy, matplotlib, seaborn). Install with: pip install -r requirements.txt

5. HVC_model_rebound_only.mat (MATLAB model file)

Example MATLAB code / parameters for running the Hodgkin–Huxley network simulations described in the paper.

Technical Details

Accessing .pickle files

To load the data in Python:

import pickle
with open('all_birds_HVC_RA.pickle', 'rb') as f:
    data = pickle.load(f)
# The data is a nested dictionary.
print(data.keys())   # top level = bird identifiers

Variable definitions (Cardinal_2.0 spreadsheets)

Every .xlsx file (one per bird) uses the same 26-column layout; each tab is a single recorded cell, and each row is one sweep/measurement. Columns are defined below; repeated variables (e.g. columns that appear twice, shown by spreadsheet software as plateau_amplitude / plateau_amplitude.1 and membrane_resistance / membrane_resistance.1) are defined once and carry the same definition and units.

times: elapsed time of the sweep relative to the start of the recording. Units: minutes (min).
SeriesR: series (access) resistance of the recording. Units: megaohms (MΩ).
MembraneR: membrane (input) resistance of the cell. Units: megaohms (MΩ).
Capacitance: whole-cell membrane capacitance. Units: picofarads (pF).
widths: action-potential width (spike half-width). Units: milliseconds (ms).
plateau_amplitude: amplitude of the post-inhibitory / plateau depolarization. Units: millivolts (mV).
half_spike_amplitude: membrane potential at the half-amplitude point of the action potential. Units: millivolts (mV).
spike_amplitudes: peak amplitude of the action potential(s). Units: millivolts (mV).
spike_baseline: membrane potential immediately preceding a spike. Units: millivolts (mV).
first_spike_amplitude: peak amplitude of the first action potential in the sweep. Units: millivolts (mV).
spike_number: number of action potentials fired in the sweep.
membrane_resistance: membrane (input) resistance computed from the voltage response to a current step. Units: megaohms (MΩ).
sag_ratios: voltage "sag" during a hyperpolarizing current step (steady-state vs peak hyperpolarization), an index of I_h. Units: ratio.
pos_command: amplitude of the depolarizing (positive) injected-current step on the sweep. Units: picoamperes (pA).
neg_command: amplitude of the hyperpolarizing (negative) injected-current step on the sweep. Units: picoamperes (pA).
holding_current: current required to hold the cell at its target membrane potential. Units: picoamperes (pA).
small_step: membrane-potential response to a small current step (used to estimate passive properties). Units: millivolts (mV).
baseline: resting/baseline membrane potential of the sweep before stimulation. Units: millivolts (mV).
last_spike_baseline: membrane potential immediately after the last action potential in the sweep. Units: millivolts (mV).
firing_frequency: firing rate evoked by the depolarizing current step. Units: hertz (Hz).
first_interspike_int: time interval between the first and second action potentials in the sweep. Units: seconds (s).
stimulated: flag indicating whether a stimulus/current step was applied on the sweep (True/False). Units: boolean.
RMP: resting membrane potential of the cell. Units: millivolts (mV).

Methodological Summary

Subjects: adult male zebra finches.
Recording type: whole-cell patch-clamp recordings in brain slices.
Units used throughout: voltage (mV), capacitance (pF), firing rate (Hz), current (pA), resistance (MΩ), time (ms / date-time).

Data Disclaimer

These data were compiled in 2022. For specific questions regarding the internal dictionary structure of the .pickle files beyond the summary statistics provided in the Cardinal_2.0 spreadsheets, please contact the corresponding author.