Deep Spatial Genomics

NOTE: now mostly serving as a template for website documentation and scratch space for random deep learning with spatial genomics data in pytorch projects

Pytorch implementations of deep learning methods for spatial genomic data.

This project is intended to harmonize tasks for published methods and speed-up deep learning-based spatial genomics methods development.

For detailed documentation see the project website

Install

pip install deep-spatial-genomics

Note: If you are running into installation issues due to the scanpy dependency you can install with Conda via the env.yaml file. Note that if your version of CUDA is not x.x you will need to specify the correct cudatoolkit version in the env.yaml file.

conda env create -f env.yaml
conda activate deep-spatial-genomics

Most tools also have google colab notebooks.

Tools

• Tangram

Tasks

• Voxel (spot) deconvolution
  • Tools: Tangram
• Mapping single cell data to spatial coordinates
  • Tools: Tangram
• Smoothing/imputation of spatial genomic data
  • Tools: Tangram
• Embedding single cell and spatial genomic data
• Voxel-Voxel interactions
• Cell-Cell interactions

Tangram

Blurb

Paper

Original repository

Detailed documentation

Colab notebook

Tangram usage

example code

Citations

@article{biancalani2021deep,
  title={Deep learning and alignment of spatially resolved single-cell transcriptomes with Tangram},
  author={Biancalani, Tommaso and Scalia, Gabriele and Buffoni, Lorenzo and Avasthi, Raghav and Lu, Ziqing and Sanger, Aman and Tokcan, Neriman and Vanderburg, Charles R and Segerstolpe, {\AA}sa and Zhang, Meng and others},
  journal={Nature methods},
  volume={18},
  number={11},
  pages={1352--1362},
  year={2021},
  publisher={Nature Publishing Group}
}