Pytorch implementation and tutorial for SAILERX.
Clone the repository.
git clone https://github.com/uci-cbcl/SAILER.git
Navigate to the root of this repo and setup the conda environment.
conda env create -f deepatac.yml
Activate conda environment.
conda activate deepatac
Please download data here and setup your data folder as the following structure:
SAILERX
|___data
|___...
For full description, please see
python train.py -h
To train with one multimodal sc-deq data (scRNA-seq + scATAC-seq). Using PBMC 10k as an example.
python train.py -d pbmc10k -cuda 0 --pos_w 20
To train with multiple multimodal sc-deq data (scRNA-seq + scATAC-seq). Using PBMC 10k + 3k as an example.
python train.py -d pbmc_batch -cuda 0 --pos_w 20 -batch True
To train with multiple multimodal sc-deq data (scRNA-seq + scATAC-seq). Using PBMC 10k + 3k as an example.
python train.py -d pbmc_hybrid -cuda 0 --pos_w 20 -batch True -t hybrid
To generate the embedding, please use the following command to load a specific model you want, subsititute PATH_TO_CKPT to the path to the ckpt you want to use (i.e., ./models/main/398.pt
).
python eval.py -d pbmc10k --name main -cuda 0 -l PATH_TO_CKPT
For more info, please use
python train.py -h
Or see evaluation examples here.