README.md

scCube v2.0.1

Simulating multiple variability in spatially resolved transcriptomics

scCube is a Python package for independent, reproducible, and platform-diverse simulation of spatially-resolved transcriptomic data

Major updates in v2.0.0

1. scCube now allows users to consider the heterogeneity within cell types and generate the spatial patterns of cell subtypes flexibly:
   • Simulation of the variability of spatial patterns within cell types
2. scCube now allows users to generate more interpretable spatial patterns in a customized manner:
   • Simulation of biologically interpretable spatial basis patterns
   • Simulation of highly customized complex spatial patterns
   • Demonstration on simulating the tumor-immune microenvironment of TNBC with three archetypical subtypes

Requirements and Installation

Create and activate Python environment

For scCube, the Python version need is over 3.8. If you have installed Python3.6 or Python3.7, consider installing Anaconda, and then you can create a new environment.

conda create -n sccube python=3.8
conda activate sccube

Install pytorch

The version of pytorch should be suitable to the CUDA version of your machine. You can find the appropriate version on the PyTorch website. Here is an example with CUDA11.6:

pip install torch --extra-index-url https://download.pytorch.org/whl/cu116

Install other requirements

cd scCube-main
pip install -r requirements.txt

Install scCube

python setup.py build
python setup.py install

Quick Start (Training new models directly)

scCube requires a data file (gene expression profiles) and a meta file (cell type/domain annotations) as the input, which can be stored as either .csv (read by pandas) or .h5ad (loaded by scanpy) formats. We have included two toy datasets in the tutorial/demo_data folder of this repository as examples to show how to use scCube.

Reference-based simulation:

• Demonstration of scCube on simulating mouse hypothalamus MERFISH data (same as the spatial reference)
• Demonstration of scCube on simulating mouse hypothalamus MERFISH data (different from the spatial reference)

Reference-free simulation:

• Demonstration of scCube on generating simulated spot-based SRT data
• Demonstration of scCube on generating simulated imaging-based SRT data

For more details about the format of input and the description of parameters, see here.

Trained models and datasets

In the current version, scCube includes about 300 trained models of various tissues derived from four human and mouse scRNA-seq atlas (Tabula Muris, Tabula Sapiens, MCA, and HCL) as well as high-quality SRT datasets. Detailed information about these models and datasets can be found here.

Tutorials (loading trained models provided by scCube)

Additional step-by-step tutorials are now available! Users can employ the trained models conveniently through the Python package to generate the new gene expression profiles of specific tissues. We provide the following tutorials as examples:

Reference-based simulation:

• Using scCube to simulate the mouse hypothalamus MERFISH data
• Using scCube to simulate the human DLPFC 10x Visium data
• Using scCube to simulate the mouse cortex STARmap data
• Using scCube to simulate the human breast cancer ST data

Reference-free simulation:

• Using scCube to simulate diverse spatial patterns
• Using scCube to simulate 3D spatial patterns
• Using scCube to simulate spatial patterns for user-specified cell types
• Using scCube to simulate the variability of resolution and spot arrangement of spot-based SRT data

About

scCube was developed by Jingyang Qian. Should you have any questions, please contact Jingyang Qian at qianjingyang@zju.edu.cn.

References

Qian, J., Bao, H., Shao, X. et al. Simulating multiple variability in spatially resolved transcriptomics with scCube. Nat Commun 15, 5021 (2024). https://doi.org/10.1038/s41467-024-49445-0