CellNavi

CellNavi is a deep learning framework designed to predict genes driving cellular transitions.

Getting Started

Installation

1. Clone the repo

git clone https://github.com/DLS5-Omics/CellNavi.git

2. Create conda environment

conda create -n cellnavi python=3.12
conda activate cellnavi

3. Install python dependencies

cd ./CellNavi
pip install -r requirements.txt

Tutorials

Please refer to tutorials/README.md and tutorials/Tutorial_perturbation.ipynb.

Datasets provided for tutorial

The full descriptions of the datasets and the studies of origin can be found in the manuscript. Here we provide the links to the pretrained model and the example datasets.

Pretrained model

• Pretrained model checkpoint and gene2token file: link

Example datasets

• Example training and testing datasets and model path file under 1,000 training step, with adjacency matrix graph and shortest path graph provided: link

Citation

Wang, T., Pan, Y., Ju, F., Zheng, S., Liu, C., Min, Y., Liu, X., Xia, H., Liu, G., Liu, H., & Deng, P. (2024). Directing cellular transitions on gene graph-enhanced cell state manifold.