zhangjx1996 / cytoset

Deep sets for multiple profiled single-cell samples

CytoSet

Introduction

CytoSet is a deep-learning based method used for predicting clinical outcome from cytometry data.

Installation

Requirements

• Python >= 3.6
• CUDA >= 10.1

pip install -r requirements.txt

Datasets

AML dataset is from https://flowrepository.org/id/FR-FCM-ZZYA. HEUvsUE dataset is from https://flowrepository.org/id/FR-FCM-ZZZU. ICS dataset is from https://flowrepository.org/id/FR-FCM-ZZZV. NK_cell dataset is from the repository of CellCNN.

The pre-processed dataset for training the model can be downloaded from the google drive.

Reproducing Results

Training

• Download pre-processed the datasets (see Datasets Section) and unpack them.
• In scripts/train/train_[Dataset].sh, set bin_file to the path of train.py and gpu to the gpu id.
• Start training: bash train_[Dataset].sh

Testing

• We provide our pre-trained model on HVTN dataset and test dataset in checkpoints.
• We also provide our model configuration for each dataset in config/model.
• To run the testing, you can use the following command:

python test.py --model checkpoints/HVTN_model.pt --config config/model/ICS/config.json --test_pkl checkpoints/test_sample.pkl

The evaluation results are:

Accuracy	Area Under Curve
0.958	0.962

Citing

@inproceedings{
    10.1145/3459930.3469529,
    author = {Yi, Haidong and Stanley, Natalie},
    title = {CytoSet: Predicting Clinical Outcomes via Set-Modeling of Cytometry Data},
    year = {2021},
    publisher = {Association for Computing Machinery},
    address = {New York, NY, USA},
    url = {https://doi.org/10.1145/3459930.3469529}
}

Contact

If you have any questions, please feel free to contact Haidong Yi ([email protected]) or push an issue on Issues Dashboard.