[new release] Conductance-based neuron dataset generator and pre-generated spike-train datasets

I am happy to announce the public release of the conductance-based neuron dataset generator, a Python toolkit for simulating large populations of conductance-based neuron (CBM) models and collecting their spike trains. This toolkit was developed to support the work presented in our preprint Fast reconstruction of degenerate populations of conductance-based neuron models from spike times and future works, and we are now making it available as a standalone open-source tool.

The repository is available on GitHub: julienbrandoit/conductance-based-neuron-dataset-generator

Together with the code, six ready-to-use spike-train datasets produced by this toolkit are publicly available on Zenodo:

Brandoit, J., Ernst, D., Drion, G., & Fyon, A. (2025). Spike-Train Datasets from Conductance-Based Neuron Models [Data set]. Zenodo. https://doi.org/10.5281/zenodo.16912160

What does the toolkit do?

The toolkit simulates large populations of conductance-based neuron models and records their spike trains (lists of spike times). Two neuron models are included:

Model	Neuron type	Ion channels
STG (Liu et al., 1998)	Stomatogastric ganglion	Na, Kd, CaT, CaS, KCa, A, H, leak
DA (Qian et al., 2014)	Midbrain dopamine	Na, Kd, CaL, CaN, ERG, NMDA, leak

Simulations support multiprocessing and optional noisy current injection (band-limited Gaussian noise), and the pipeline is designed to scale to HPC clusters via Slurm array jobs.

Two conductance sampling strategies

A key feature of the toolkit is that conductance sets can be generated by two different strategies:

• Monte Carlo — each maximal conductance is sampled independently from a uniform or Gamma distribution over biologically plausible ranges. This provides a broad, unconstrained exploration of conductance space.
• DICs framework — sampling is guided by Dynamic Input Conductances (DICs) to generate degenerate populations: groups of neurons that exhibit the same activity pattern despite having widely different individual conductance values. This is the same framework used in our preprint.

The DICs framework is grounded in the following works:

• Drion G. et al. Dynamic Input Conductances Shape Neuronal Spiking. eNeuro 2(1), 2015. doi:10.1523/ENEURO.0031-14.2015
• Fyon A. et al. Dimensionality reduction of neuronal degeneracy reveals two interfering physiological mechanisms. PNAS Nexus, 3(10), pgae415, 2024. doi:10.1093/pnasnexus/pgae415
• Brandoit J. et al. Fast reconstruction of degenerate populations of conductance-based neuron models from spike times. arXiv:2509.12783, 2025. https://arxiv.org/abs/2509.12783

Pre-generated datasets

Six datasets (version 2.0) are immediately available on Zenodo, spanning the two neuron models, both sampling strategies, and different noise conditions:

Dataset	Model	Sampling	Noise (σ)	Splits	Total rows
dics_stg_nonoise	STG	DICs	—	train / val / test	1 583 999
dics_stg_noise_5	STG	DICs	5 µA/cm²	train / val	1 197 849
dics_da_nonoise	DA	DICs	—	train / val	1 682 276
dics_da_noise_5	DA	DICs	5 µA/cm²	train / val	279 780
mc_gamma_stg_noise_5	STG	Monte Carlo (Gamma)	5 µA/cm²	train / val	994 500
mc_uniform_stg_noise_3	STG	Monte Carlo (Uniform)	3 µA/cm²	train / val	743 500

Each dataset entry is documented in detail in the repository (ion channels, DICs parameters, simulation window, noise settings, and available splits).

Note: A version 1.0 was released earlier in 2025 as an early, poorly documented snapshot. Version 2.0 supersedes it and should be used for all new work.

Citation

If you use the toolkit in your research, please cite our preprint:

@misc{brandoit2025fast,
  title        = {Fast reconstruction of degenerate populations of conductance-based neuron models from spike times},
  author       = {Brandoit, Julien and Ernst, Damien and Drion, Guillaume and Fyon, Arthur},
  year         = {2025},
  eprint       = {2509.12783},
  archivePrefix = {arXiv},
  primaryClass = {q-bio.NC},
  url          = {https://arxiv.org/abs/2509.12783}
}

If you also use one of the pre-generated datasets, please additionally cite the data record:

@dataset{brandoit2025dataset,
  author    = {Brandoit, Julien and Ernst, Damien and Drion, Guillaume and Fyon, Arthur},
  title     = {Spike-Train Datasets from Conductance-Based Neuron Models},
  year      = {2025},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.16912160},
  url       = {https://doi.org/10.5281/zenodo.16912160}
}

Contact: For questions or collaborations, please reach out to me at jbrandoit@uliege.be.

Damien Ernst’s webpage: Damien Ernst

Guillaume Drion’s webpage: Guillaume Drion

Arthur Fyon’s webpage: Arthur Fyon