Orthogonal graph autoencoders for constructing cell-cell communication networks on spatial transcriptomics data.
OrgaCCC reconstructs cell-cell and cell-type communication networks using the orthogonal graph autoencoders at both the cell/spot and gene levels. Through clustering analysis, it refines cell classification and uncovers developmental trajectories, as well as spatial and functional heterogeneity of cells. Additionally, OrgaCCC employs gene sensitivity analysis to identify key genes involved in cell communication, offering valuable insights into critical molecular interactions.
OrgaCCC is implemented in python 3.11.3.
OrgaCCC depends on the following packages: matplotlib, numpy, pandas, scanpy, scipy, tensorflow. See dependency versions in requirements.txt.Installation of the dependencies may take several minutes.
pip install --requirement requirements.txt
The input data should be an AnnData object (adata) with the following structure:
- Gene Expression Matrix: Cells in rows and genes in columns.
adata.obs['celltype']: A column containing cell type annotations for each cell.adata.obsm['spatial']: A two-column array storing the spatial coordinates(x, y)of each cell.
Example:
adata = sc.read_h5ad(r"data/cortex.h5ad")
If your file is located elsewhere, you will need to manually update the path in main.py
We provide a preprocessing module to transform the spatial coordinates of single-cell transcriptomics data into an adjacency matrix.
Run the script:
python main.py
The final outputs the model's AUC performance, a cell-cell communication matrix, a cell type communication matrix, and results from clustering, UMAP visualization, and PAGA analysis, providing a comprehensive view of cell interactions and relationships.