DeepGSEA: Explainable Deep Gene Set Enrichment Analysis for Single-cell Transcriptomic Data

DeepGSEA is a deep learning-enhanced gene set enrichment (GSE) analysis method which leverages the expressiveness of interpretable, prototype-based neural networks to provide an in-depth analysis of GSE.

Prerequisite

Install all required packages in ./requirements.txt (tested on Python 3.9.6)

pip install -r requirements.txt

Quick start

Reproduce results on the glioblastoma data

1. Downlaod gene set databases from MSigDB following ./data/msigdb/download.txt

2. Download the scRNA-seq data following ./data/GSE132172/download.txt

3. Move to the directory ./deepgsea

   cd ./deepgsea
   

4. Run DeepGSEA on the dataset with interpretations

   sh ./scripts/run_glioblastoma.sh
   

5. Run the sigificance test on the results

   python pvalue_real.py --data glioblastoma
   

Citation

If you find our research useful, please consider citing:

@article{xiong2024deepgsea,
  title={DeepGSEA: Explainable Deep Gene Set Enrichment Analysis for Single-cell Transcriptomic Data},
  author={Xiong, Guangzhi and John LeRoy, Nathaniel and Bekiranov, Stefan and Sheffield, Nathan and Zhang, Aidong},
  journal={Bioinformatics},
  pages={btae434},
  year={2024},
  publisher={Oxford University Press}
}