Out-of-distribution prediction with disentangled representations for single-cell RNA sequencing data
The pdf for the paper can be found here.
Reproducing results from Jupyter Notebooks:
- JN1_train.ipynb - loads, normalizes and trains the data
- JN2_kang_analysis.ipynb - further analyses the trained model for Kang dataset. It calculates both types of disentanglement scores and also visualizes the latent space as well as gene-feature space plots.
- JN3_dentate_analysis.ipynb - further analyses the trained model for Dentate Gyrus dataset.
- JN4_Out of Distribution Prediction.ipynb - implements OOD Prediction and analyses the results further
- Model Comparisons folder provides the notebooks to reproduce the model architecture comparison plots.
- The environment.yml file specifies the conda environment in which the project was run.
- Data can be accessed from here
please consider citing
@inproceedings{lotfollahi2020out,
title={Out-of-distribution prediction with disentangled representations for single-cell RNA sequencing data},
author={Lotfollahi, Mohammad and Dony, Leander and Agarwala, Harshita and Theis, Fabian},
booktitle={ICML 2020 Workshop on Computational Biology (WCB) Proceedings Paper},
volume={37},
year={2020}
}