SAHA MaxFuse

This repository contains the RNA-protein integration of cell type annotations for the SAHA project.

PreprocessPRT.ipynb

This script takes the protein Anndata object and splits it into individual organs to prepare for integration. Furthermore, it clusters each individual organ protein Anndata.

PreprocessRNA.ipynb

This script takes the RNA Anndata object and splits it into individual organs to prepare for integration.

MaxFuse.py

This script performs the cell matching using the MaxFuse package. Before running the script, there are a few things to edit.

In line 19, change the obs_column to the selected cluster obs name on the protein Anndata object that encompasses the number of cell types from the RNA annotations.

obs_column = 'leiden_3'

In line 22-23, change how Anndata objects for protein and RNA are named so that they can be read into the script.

adata_PR = ad.read_h5ad(f'PRT_{tissue}.h5ad')
adata_RNA = ad.read_h5ad(f'RNA_{tissue}.h5ad')

In line 33, change what the RNA annotation obs column is named.

labels_rna = adata_RNA.obs['Insitutype_Labelled'].to_numpy()

In line 39, change what the csv file is called that contains the analogous protein-RNA name matching.

correspondence = pd.read_csv('./protein_gene_conversion.csv')

Run the script like so and it will generate a matching_{tissue}_{obs_column}.csv file.

python MaxFuse.py "$tissue"

PostMaxFuse.ipynb

This script takes the matching and assigns cell typing to the protein Anndata object under the obs column name 'celltype_mf'. You can check how the integration performed by examining the cell expression dot plot.

And the UMAP.

PostMaxFuseGrannular.ipynb

This script is similar to the one above but instead of integrating broad cell types, we integrated granular cell types plotted the results.

BigUMAP.ipynb

This script combines all the individually matched organs and combine together for analysis. We generated the cell expression heatmap from this script.

PrettyPictures.ipynb

This script generates the spatial mapping of the cell types. First, we figure out what the fovs so we can select regions to plot and compare.

From this spatial information, we can examine similar regions in protein and RNA and compare how well the integration performed.