A unified framework for extracting single-cell information from high-resolution spatial transcriptomics
CellART is a unified framework for extracting single-cell information from high-resolution ST data. The primary objectives are to accurately delineate boundaries for individual cells and further annotate their cell types. By integrating deep neural networks with probabilistic models, CellART leverages multimodal data, including spatial transcriptomics, staining images, and scRNA-seq references, to perform simultaneous cell segmentation and cell type annotation, thereby optimizing the analytical process.
Visit our documentation for installation, examples and reproducing the results in our paper. All the data files for reproducing the result are deposited at Zenodo.
CellART can be installed via two approaches: using pip or directly from the GitHub repository.
You can install the stable release of CellART directly from pip. First, ensure you have set up a compatible Python environment using conda or any other package manager. Then, execute the following commands:
$ conda create -n cellart python=3.10
$ conda activate cellart
$ pip install cellartNOTE: Due to differences in GPU models and CUDA versions, you may need to manually reinstall PyTorch and Tensorflow to ensure compatibility. CellART relies heavily on GPU acceleration for efficient processing of large-scale spatial transcriptomics datasets.
# Clone the repository
$ git clone https://github.com/YangLabHKUST/CellART.git
# Navigate to the cloned directory
$ cd CellART
# Install the package
$ pip install .- Segment and annotate Xenium CRC dataset from raw data
- Segment and annotate VisiumHD CRC dataset from raw data
CellART is currently under review.
The python package CellART is developed and maintained by Yuheng Chen.
Please feel free to contact Yuheng Chen, Prof. Jiashun Xiao or Prof. Can Yang if any inquiries.