We processed the embeddings of all data used in the paper and saved them. You need to download the gene expression data for reproducing the baseline results from Figshare: https://doi.org/10.6084/m9.figshare.24049200 . Then you need to used the split_data_SCAD_5fold_norm.py to generate the split data for 5-fold cross validation.
cd /data/split_norm/
## with outembedding
python split_data_SCAD_5fold_norm.py --drug Sorafenib --emb 0
## with embedding
python split_data_SCAD_5fold_norm.py --drug Sorafenib --emb 1
The directroy with drug name such as split_norm/Sorafenib/ will be created. You can change the druge name --drug to generate other drugs' split data.
# cd to the SCAD folder
# Sorafenib
## without embedding
CUDA_VISIBLE_DEVICES=1 python model/SCAD_train_binarized_5folds-pub.py -e FX -d Sorafenib -g _norm -s 42 -h_dim 512 -z_dim 128 -ep 20 -la1 5 -mbS 8 -mbT 8 -emb 0
## with embedding
CUDA_VISIBLE_DEVICES=1 python model/SCAD_train_binarized_5folds-pub.py -e FX -d Sorafenib -g _norm -s 42 -h_dim 1024 -z_dim 256 -ep 80 -la1 0.2 -mbS 32 -mbT 32 -emb 1
Follow the command in run.sh for more results. It will take several minutes to train the model.
| Drug | Method | Single Cell AUROC |
|---|---|---|
| Sorafenib | Baseline | 0.56 |
| Sorafenib | scFoundation | 0.84 |
| NVP-TAE684 | Baseline | 0.62 |
| NVP-TAE684 | scFoundation | 0.84 |
| PLX4720_451Lu | Baseline | 0.38 |
| PLX4720_451Lu | scFoundation | 0.66 |
| Etoposide | Baseline | 0.66 |
| Etoposide | scFoundation | 0.68 |
scanpy
pandas
pytorch
scikit-learn
The plot-publish.ipynb is the file of reproducing results. If you want to run this notebook by your self, you can download the processed data from figshare: https://dx.doi.org/10.6084/m9.figshare.24049200 .