NeuPRINT

Official implementation of Neuronal Time-Invariant Representations (NeuPRINT).

Learning Time-Invariant Representations for Individual Neurons from Population Dynamics. NeurIPS 2023

Lu Mi*, Trung Le*, Tianxing He, Eli Shlizerman, Uygar Sümbül

Datasets

Our framework is evaluated on the datasets from Bugeon et al. 2022, Nature (A transcriptomic axis predicts state modulation of cortical interneurons), download dataset from this link.

Environment Setup

Assuming you have Python 3.8+ and Miniconda installed, run the following to set up the environment with necessary dependencies:

conda env create -f environment.yml

Run Experiments

Multiple data split and evaluation modes you can test with this repo with modification on the main.py file:

1. Single animal: In our paper where we reported on Table 1 from a single animal (SB25) with train/val/test neuron split. Please check our description “We first evaluate our approach on one animal (SB025) across 6 sessions. The recordings from this animal include 2481 neurons in total.
2. Multi-animal: Table 2 refers to multi-animal setting, where all mice are included during the training, with some neurons are heldout for evaluations. Please check the description "We then extend our analysis on functional recordings from 4 mice (SB025, SB026, SB028, SB030) across 17 sessions." "We introduce a downstream classification task to predict the subclass label with supervised learning, where the neurons with subclass labels from all sessions are randomly split into train, validation and test neurons with a proportion of 80% : 10% : 10%."
3. Cross-animal: The most challenging setting with cross-animal training/val/test split, where our dynamical model f is trained on three mice SB25, SB26, SB30, and for the held-out mouse SB28 for evaluation, we only finetune \phi with f fixed, and the subclass classifier is only trained on SB25/SB26/SB30.

python main.py --exp-tag neuprint_train

Citations

If you find our code helpful, please cite our paper:

@article{mi2024learning,
  title={Learning Time-Invariant Representations for Individual Neurons from Population Dynamics},
  author={Mi, Lu and Le, Trung and He, Tianxing and Shlizerman, Eli and S{\"u}mb{\"u}l, Uygar},
  journal={Advances in Neural Information Processing Systems},
  volume={36},
  year={2024}
}

If you use the dataset, please cite this paper:

@article{bugeon2022transcriptomic,
  title={A transcriptomic axis predicts state modulation of cortical interneurons},
  author={Bugeon, Stephane and Duffield, Joshua and Dipoppa, Mario and Ritoux, Anne and Prankerd, Isabelle and Nicoloutsopoulos, Dimitris and Orme, David and Shinn, Maxwell and Peng, Han and Forrest, Hamish and others},
  journal={Nature},
  volume={607},
  number={7918},
  pages={330--338},
  year={2022},
  publisher={Nature Publishing Group UK London}
}