Unbalanced Mapping Approach for High-Resolution Spatial Transcriptomics
The notebook main_annotate.ipynb shows how to annotate High-Resolution Spatial Transcriptomics data with high accuracy using a reference dataset. The following is the structure of the package:
src/ot_annotator.py: Used for subclustering and OT mapping.src/plot_hp.py: Used to plot hyperparameters scatterplot.src/compare_viz.py: Used to evaluate results (comparison to other methods and visualization).main_annotate.ipynb: example ...
The following are the main objects (useful outputs) of the annotator class:
annotator.T: The transport plan.annotator.X_reconstructed: the reconstructed matrix.annotator.adata: the updated ST target modality with annotator.adata.obs['predicted_annotation'] is the transferred annotation and annotator.adata.obs['central_cell_membership'] is the subclustering labeled with the central cell of the subcluster (the same is in annotator.adata_ref).selected_central_cells: the subclusters that are annotated with high probability based on T. The threshold is selected based on the elbow approach.