readme.md

Download data

The following datasets will be used for testing normalization methods. For each only the data from lung will be used.

1. LungMAP — Human data from a broad age healthy donor group.
2. Tabula Sapiens - All Cells
3. COVID-19 immune features revealed by a large-scale single-cell transcriptome atlas

bash scripts/download/download_data.sh

Process data

Select only lung data and 10X

bash ./scripts/processing/subsample_lung.sh

Remove cells with <500 expressed genes and make raw data adata.X

bash ./scripts/processing/drop_low_coverage_cells.sh  

Sub-sample to 100K max cells

This will subsample cells to a max of 100K cell per dataset, the sampling is done per cell type so that all cell tpyes are preserved with the same original proportions

 bash ./scripts/processing/subsample_h5ad.sh

Normalize data

Do quantile normalization

 bash ./scripts/normalizing/rankit.sh

Do iNMF-like normalization normalization

bash ./scripts/normalizing/inmf.sh

Do Z-scores

bash ./scripts/normalizing/z_scores.sh

Do scvi normalization

mkdir -p ./data/normalized/scvi

python3 ./scripts/normalizing/scvi.py

Get marker genes

mkdir -p ./data/marker_genes/
python3 ./scripts/processing/find_marker_genes.py rankit ./data/normalized/scvi/all.h5ad ./data/marker_genes/marker_genes_rankit.tsv
python3 ./scripts/processing/find_marker_genes.py scvi ./data/normalized/scvi/all.h5ad ./data/marker_genes/marker_genes_scvi.tsv

Getting data ready for R notebook

Save scvi normalization of data only from lung map

mkdir -p ./data/for_notebook/
python3 ./scripts/processing/save_mtx_obs_var.py all ./data/normalized/scvi/lung_map.h5ad ./data/for_notebook/lung_map

Save all normalization methods of all datasets

mkdir -p ./data/for_notebook/
python3 ./scripts/processing/save_mtx_obs_var.py all ./data/normalized/scvi/all.h5ad ./data/for_notebook/all_datasets