This repository contains notebooks and supporting materials to reproduce analyses from the manuscript:
Identification of cell types, states and programs by learning gene set representations
It is designed to complement the main software package scDECAF by providing end-to-end, executable examples.
For installation instructions and environment setup, please refer to the main scDECAF repository.
The following Jupyter notebooks illustrate how to reproduce key results and analyses:
- Pathway/gene signature screening and scoring Kang et al. (2018): 25K PBMC single cells
- Optimization of sparsity operator Experimentation with the shrinkage penalty and gene set screening results in Kang et al. (2018)
- PMBC COVID-19 analysis Combining reference atlas mapping and Milo analysis with scDECAF gene set screening
- Drug2cell analysis with scDECAF [Running scDECAF with pre-computed Drug2cell scores in HECOA Organoid Atlas]
More examples will be added as the manuscript evolves.
-
Clone this repository:
git clone https://github.com/DavisLaboratory/scDECAF-reproducibility.git cd scDECAF-reproducibility -
Ensure that you have installed
scDECAFand its dependencies (see installation guide). -
Open the notebooks with Jupyter Lab or Jupyter Notebook:
jupyter lab
-
Run through the cells to reproduce figures and results.
Data required for running the notebooks can be accessed from the sources cited in the paper (e.g., Kang et al. 2018). Where possible, direct download links are provided inside the notebooks.
If you use these materials, please cite:
Hediyehzadeh, Whitfield et al., Identification of cell types, states and programs by learning gene set representations, bioRxiv (2023). DOI: 10.1101/2023.09.08.556842
This repository follows the same license as the main scDECAF repository.