superot

This is the code repository for the paper https://arxiv.org/pdf/2007.12098.pdf. The dataset can be obtained from Experiment 1 in https://github.com/AllonKleinLab/paper-data/blob/master/Lineage_tracing_on_transcriptional_landscapes_links_state_to_fate_during_differentiation/README.md.

Here is a description of each of the files. For data_preprocess.py, main_conditional.py, main.py, and main_supervised.py, please change COUNTS_MATRIX, CLONE_ANNOTATION, CELL_METADATA to the corresponding file names in your directory.

1. GAN.py --> GAN model code
2. cGAN.py --> Conditional GAN model code
3. data_preprocess.py --> Code for setting up the data loaders in the unsupervised, semi-supervised, and supervised settings.
4. utils.py --> Code for parsing arguments / data loaders
5. main_conditional.py --> Code for training and evaluating the conditional GAN in the unsupervised setting.
6. main.py --> Code for training and evaluating the GAN in the semi-supervised setting. Note that, depending on the number of supervised points you choose to use, you will have to change 'num_points' accordingly.
7. main_supervised.py --> Code for training and evaluating the GAN in the supervised setting.

To use, first save the necessary data loaders by running data_preprocess.py, being sure to download all the data and have it saved in the same directory first. Then, create an empty folder named 'results' and start up a new visdom server by running python -m visdom.server -port 3000 in terminal. In a new tab, run either main_conditional.py, main.py or main_supervised.py to obtain your results.