scMVP is a python toolkit for joint profiling of scRNA and scATAC data and analysis with multi-modal self-attention generation model.

• 20220815
  Complete the test for GPU version on NVIDA 3090 platform with CUDA version 11.2.
  Add scRNA and scATAC input check in tutorial.
  

Environment requirements:
scMVP requires Python3.7.x and Pytorch.
For example, use miniconda to install python and pytorch of CPU or GPU version.
We have tested the GPU version on NVIDA 1080Ti platform with CUDA version 10.2.

conda install -c pytorch python=3.7 pytorch
# if you do not have jupyter notebook/ipython notebook, you can also install by conda
conda install jupyter

Then you can install scMVP from github repo:

# first move to your target directory
git clone https://github.com/bm2-lab/scMVP.git
cd scMVP/
python setup.py install

Try import scMVP in your python console and start your first tutorial with scMVP!

Jupyter notebooks for other datasets analyzed and benchmarked in our GB paper are deposited in folder

Download link: baidu cloud disk

• update 23/10/22 : google drive link for cellline datasets: google drive
  

Download code: mkij

• pre_trainer.pkl scRNA pretraining models
  
• pre_atac_trainer.pkl scATAC pretraining models
  
• multi_vae_trainer.pkl scMVP training models
  

Applying scMVP to sci-CAR cell line mixture. demo

• Training and visualization with scMVP.
• Pretraining and transferring to scMVP(perform better in large dataset).

A deep generative model for multi-view profiling of single-cell RNA-seq and ATAC-seq data. Genome Biology 2022 paper

Prof. Qi Liu: qiliu@tongji.edu.cn
Dr. Gaoyang Li: lgyzngc@gmail.com
Shaliu Fu: adam.tongji@gmail.com