Consistent cross-modal identification of cortical neurons with coupled autoencoders
@article{gala2021consistent,
title={Consistent cross-modal identification of cortical neurons with coupled autoencoders},
author={Gala, Rohan and Budzillo, Agata and Baftizadeh, Fahimeh and Miller, Jeremy and Gouwens, Nathan and Arkhipov, Anton and Murphy, Gabe and Tasic, Bosiljka and Zeng, Hongkui and Hawrylycz, Michael and S{\"u}mb{\"u}l, Uygar},
journal={Nature Computational Science},
volume={1},
number={2},
pages={120--127},
year={2021},
publisher={Nature Publishing Group}
}
Consistent identification of neurons in different experimental modalities is a key problem in neuroscience. While methods to perform multimodal measurements in the same set of single neurons have become available, parsing complex relationships across different modalities to uncover neuronal identity is a growing challenge. Here, we present an optimization framework to learn coordinated representations of multimodal data, and apply it to a large multimodal dataset profiling mouse cortical interneurons. Our approach reveals strong alignment between transcriptomic and electrophysiological characterizations, enables accurate cross-modal data prediction, and identifies cell types that are consistent across modalities.
data/proc/
contains the processed dataset used for Gala et al. 2021.notebooks/data_proc_T.ipynb
and notebooks/data_proc_E.ipynb
for pre-processing steps. conda
environment, and install depencies (see requirements.yml
). The models can be run with tensorflow
versions 2.1
to 2.5
setup.py
in this reposiory, and use pip install -e .
cplAE_TE/train.py
to start training a model.You can also play around with a minimal version of the coupled autoencoders code (see minimal
folder in this repository) hosted on a cloud environment at CodeOcean.
The main points covered by earlier work: